GPR Annan 2003

8/20/2019 GPR Annan 2003

http://slidepdf.com/reader/full/gpr-annan-2003 1/286

Copyright 2003 Sensors & Software Inc. 03-0003-00

Ground Penetrating Radar

Applications

Principles, Procedures

A P Annan

8/20/2019 GPR Annan 2003


Copyright 2003 Sensors & Software Inc.

Copies of these notes are available on CD in PDF format from Sensors & Software Inc. for a fee of $50.00

plus shipping and handling. If you wish to obtain copies of these notes, please fill in the information indi-

cated here and send with a cheque, money order or credit card information to Sensors & Software Inc.

PLEASE FILL OUT AND FAX BACK

Name: _______________________________________________________________________

Company: _______________________________________________________________________

Address: _______________________________________________________________________

_______________________________________________________________________

City: __________________________ Prov/State: _________ Postal/Zip Code: _________________

Phone Number: ( ) _______________________ Fax Number: ( ) ____________________

E-mail: _______________________________________________________________________

Product/Equipment Requested: Ground Penetrating Radar Principles, Procedures & Applications

METHOD OF PAYMENT

Do Not Send Cash By Mail

VISA/MC #:

NAME OF PERSON AS IT APPEARS ON CARD:

EXPIRATION DATE:

SIGNATURE:

(For Credit Card Payments Only)

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications Table of Contents

i

Table of Contents

1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 What is GPR? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Current Activities & Future Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Basic Electromagnetic Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.3 Constitutive Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4 Wave Nature of EM Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.5 Low Loss – Damped EM Wave Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.6 Sinusoidally Time Varying Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.7 Electromagnetic Impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.8 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.9 EM Fields From Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.10 Far Field Approximation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.11 Antenna Patterns, Directivity, Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3 Physical Properties I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.1 Why are Physical Properties Important?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2 Conduction Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.3 Displacement (Polarization) Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4 Total Current Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.5 Magnetic Permeability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.6 Complex Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.7 Electrical Properties of Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.8 Real GPR Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.8.1 DNAPL Spill . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52

3.8.2 Conductive Contaminant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .53

8/20/2019 GPR Annan 2003


Table of Contents GPR Principles, Procedures & Applications

ii

4 EM Wave Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.1 Wavefronts and Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.2 Wave Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.3 GPR Plateau . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.4 Material Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.5 Snell’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.6 Critical Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.7 Fresnel Reflection Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4.8 Thin Layer Reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.9 Plane Wave Model for Finite Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.10 GPR Source Near an Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4.11 Resolution and Zone of Influence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.12 Scattering from a Localized Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

4.13 Scattering Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4.14 Propagation Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

5 GPR Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5.1 GPR Measurement Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5.2 Fundamental Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5.3 GPR Timing and Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.3.1 Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .92

5.3.2 Resolution and Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.4 GPR Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5.5 System Elements & Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

5.6 GPR Signal Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

5.6.1 Correlation Acquisition GPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.6.2 Frequency Domain Mixer GPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .99

5.6.3 Equivalent Time sampling GPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101

5.6.4 Common Signal Capture Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

5.7 Real-Time Signal Processing Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

5.7.1 Deconvolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .103

5.7.2 GPR Wavelet Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .104

5.8 System Performance Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

http://0.0.0.0/

http://0.0.0.0/

http://0.0.0.0/

8/20/2019 GPR Annan 2003



iii

5.9 Signal Amplitude & Recording Dynamic Range. . . . . . . . . . . . . . . . . . . . . . . . 107

5.9.1 Characterizing System Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .107

5.9.2 Dynamic Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .111

5.9.3 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .114

5.10 Centre Frequency and Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

5.11 Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

5.11.1 Dipole Impulse Response or Transfer Function. . . . . . . . . . . . . . . . . . . . . . . . .118

5.11.2 Antenna Directivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .119

5.11.3 Shielding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .122

6 Modelling of GPR Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

6.1 The Purpose of Modelling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

6.2 GPR Modelling Types & Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

6.3 One-dimensional (1D) Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

6.3.1 Radar Range Equation (RRE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .128

6.3.2 One-Dimensional (1D) Layered Earth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .129

6.4 Two-Dimensional (2D)Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

6.4.1 Ray Tracing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .130

6.4.2 FK 2D Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .132

6.4.3 Full Finite Difference (FD) Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .133

6.4.4 2½ D Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .133

6.5 Three-Dimensional (3D) Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

6.5.1 3D Ray Tracing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .133

6.5.2 3D Finite Difference Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .133

6.5.3 Integral Equation – Equivalent source Scattering . . . . . . . . . . . . . . . . . . . . . . .135

6.6 Inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

7 Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

7.1 Evaluating GPR Suitability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

7.2 Reflection Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

7.2.1 Selecting Operating Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .146

7.2.2 Estimating the Time Window. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .148

7.2.3 Selecting Temporal Sampling Interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .149

7.2.4 Selecting Station Spacing (Spatial Sampling Interval) . . . . . . . . . . . . . . . . . . .149

8/20/2019 GPR Annan 2003



iv

7.2.5 Selecting Antenna Separation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .151

7.2.6 Survey Grid and Coordinate System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .152

7.2.7 Selecting Antenna Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .153

7.3 CMP/WARR Velocity Sounding Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

7.4 3D Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

7.5 Borehole Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

8 Data Processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

8.1 Data Editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

8.2 Basic Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

8.2.1 Dewow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .159

8.2.2 Time Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .159

8.2.3 Temporal and Spatial Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .163

8.2.4 Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .166

8.3 Advanced Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

8.3.1 Attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .167

8.3.2 Deconvolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .168

8.3.3 Background Subtraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .168

8.3.4 Velocity Analysis from CMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .169

8.3.5 Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .169

8.4 Visual/Interpretation Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

8.4.1 Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .170

8.4.2 Event Picking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .171

8.4.3 Volume Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .172

8.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

8.6 Final Words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

9 Interpretation, Concepts & Pitfalls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

9.1 Gradational Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

9.2 Velocity Determination Using Hyperbolic Fitting. . . . . . . . . . . . . . . . . . . . . . . 180

9.3 Polarity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

9.4 Airwave Events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

9.5 X Marks the Spot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

8/20/2019 GPR Annan 2003



v

9.6 Plastic Water Pipe Diameter Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

9.7 Antenna Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

9.8 Ringing on Radar Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

9.9 GPR Time Zero & Depth Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

9.10 GPR Antenna Elevation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

9.11 Determining Layer Thickness & Velocity from CMP Data . . . . . . . . . . . . . . . 205

9.11.1 Analysis Procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .205

9.11.2 Intercept Time & RMS Velocity Conversion to Depth & Interval Velocity . . . 209

9.11.3 Direct Air Wave & Ground Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .210

9.11.4 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .210

9.12 CMP Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

10 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

10.1 Mining & Quarrying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

10.2 Geotechnical & Environmental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

10.3 Forensic & Archeology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

10.4 Buried Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

10.5 Structure Assessment (NDT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

10.6 Military, Law Enforcement & Espionage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262

10.7 Bio Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

10.8 Snow & Ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264

10.9 Educational material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269

11 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

8/20/2019 GPR Annan 2003



vi

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 1-Introduction

1

1 INTRODUCTION

Ground penetrating radar (GPR) is a relatively new geophysical technique. The last decade has seen major advancesas the technology matures. The history of GPR is intertwined with the diverse applications of the technique. GPR hasthe most extensive set of applications of any geophysical technique leading to a wide range of application spatialscales and concomitant diversity of instrument configurations.

This document is intended to introduce the physical principles underpinning GPR and provide practical guidance tousers of the method.

1.1 WHAT IS GPR?

Before delving into the history, GPR needs definition. GPR uses electromagnetic fields to probe lossy dielectric mate-rials to detect structures and changes in material properties within the materials (Davis & Annan (1989)). Reflectionand transmission measurements, as depicted in Figure 1-1, are employed. Most applications to date have been in nat-ural geologic materials, but widespread use also occurs for man-made composites such as concrete, asphalt and other construction materials. In lossy dielectric materials, electromagnetic fields can penetrate to a limited depth before being absorbed. Hence, penetration is always an issue.

With GPR, the electromagnetic fields propagate as essentially non-dispersive waves. The signal emitted travels

through the material, is scattered and/or reflected by changes in impedance giving rise to events which appear similar to the emitted signal. In other words, signal recognition is simple because the return signal looks like the emitted sig-nal. Figure 1-2 depicts the general nature of GPR reflection profiling. Figure 1-3 shows the GPR response of two roadtunnels.

GPR field behavior occurs over a finite frequency range generally referred to as the GPR plateau where velocity andattenuation are almost frequency independent. The GPR plateau usually occurs in the 1 MHz to 1000 MHz frequencyrange. At lower frequencies the fields become diffusive in character and pulses are dispersed. At higher frequenciesseveral factors increase signal absorption such that penetration is extremely limited.

Figure: 1-1 Ground penetrating radar uses radio waves to probe the subsurface of lossy dielectric materials. Twomodes of measurement are common. In the first, detection of reflected or scattered energy is used. In the second, vari-

ation after transmission through the material is used to probe a structure.

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003


1-Introduction GPR Principles, Procedures & Applications

2

Figure: 1-2 In reflection mode, a GPR instrument is normally moved along a survey line acquiring responses at reg-ular intervals which are used to create a cross sectional image of the ground.

Figure: 1-3 GPR cross section obtained with a 50 MHz system traversed over two road tunnels.

8/20/2019 GPR Annan 2003



3

1.2 HISTORY

The following is necessarily brief and intended to give high lights. References lead to other perspectives for thoseinterested in a more extensive understanding of GPR. It is interesting to note that accounts of some activities are pub-lished many years later and sometimes not all.

1900 – 1950

During this time a great deal of research on radio wave propagation above and along the surface of the earth occurred.Although several hints at the possibility of using radio waves to probe the subsurface are mentioned, there are noreports of successfully making this type of measurement. Vast number of papers appeared on the subject of communi-cations, direction finding and radar.

1950 – 1955

In this time frame, the first reported attempt at measuring subsurface features with radio wave signals was reported.El Said (1956) attempted to use the interference between direct air transmitted signals and signals reflected from thewater table to image the water table depth.

1955 – 1960

The next reported observation of radio frequency sounding of geological materials came about when the USAFreported altimeter errors when attempting to land aircraft on the Greenland ice sheet (Waite and Schmidt (1961)).This was the first time that repeatable indications of penetration into the subsurface through a naturally occurringmaterial were reported. This spawned the era of researchers focused on developing radio echo sounding in ice.

1960 – 1965

The majority of activity during this interval involved the radio echo sounding in ice. Groups, such as the Scott Polar Research Institute at Cambridge, Bailey et al (1964) and the Geophysical and Polar Research Center at the University

of Wisconsin, Bentley (1964), Walford (1964) were active in polar regions and also on glaciers.

1965 – 1970

During this time the ice radio echo sounding activity continued. In addition, applications in other favorable geologicmaterials started to be explored. Cook (1973) explored the use in coal mines since coal can be a low loss dielectricmaterial in some instances. Similarly, Holser et al (1972), Unterberger (1978) and Thierbach (1973) initiated evalua-tions in underground salt deposits for similar reasons.

This period was the start of lunar science mission planning for the Apollo program. Several experiments were devisedto examine the lunar subsurface which was believed to have electrical character similar to that of ice. The work of Annan (1973) reports on some of these developments.

8/20/2019 GPR Annan 2003



4

Figure: 1-4 The surface electrical properties experiment carried out on Apollo 17 used a 3 component vectorreceiver mounted on the lunar rover and a dual axis multi-frequency dipolar antenna laid out on the surface to sound

the subsurface.

1970 – 1975

This period saw numerous advances. The Apollo 17 lunar exploration program involved the surface electrical proper-ties experiment (Figure 1-4) which used interferometry concepts similar to the work carried out by El Said (1956)while the work lunar orbiter carried a pulsed radar sounder similar to the ice sounders which made measurementsfrom orbit over the lunar surface (Simmons et al (1973) and Ward et al (1973)).

During the same period Morey and others formed Geophysical Survey Systems Inc. which has been manufacturingand selling ground penetrating radar since that time (Morey (1974)).

In addition a better understanding of electrical properties of geologic materials at radio frequencies started to becomeavailable. Work such as that by Olhoeft (1975) led to a much better understanding of the electrical character of naturaloccurring geological materials and the relationship between electrical conductivity and dielectric polarization of thesematerials.

1975 – 1980

During this period, applications started to grow because of the availability of technology and a better understandingof geology. The Geological Survey of Canada explored a number of applications, the primary one being a better understanding of permafrost terrain in the Canadian Arctic. A GPR system in operation is shown in Figure 1-5. Pro- posals for pipelines out of the Arctic to carry oil and gas to southern markets drove a great deal of interest in engi-neering in frozen soil and environments. GPR was a tool which offered great promise and some of the initial resultsare reported by Annan and Davis (1976).

During this period the effect of scattering on radio echo sounding in temperate glaciers became better understood.The impact of scattering and the need for lower frequency radars was reported by Watts and England (1976).

Experiments with GPR were reported by the Stanford Research Institute where measurements were made by Dolphinet al (1978) for archeological applications.

Other work carried out in this period which paralleled the Geological Survey of Canada permafrost efforts was lead by Olhoeft at the United States Geological Survey who worked on the Alaska pipeline routes.

8/20/2019 GPR Annan 2003



5

Figure: 1-5 GPR system being used to survey potential pipeline routes in the Canadian Arctic (1975).

Extensive work was carried out in potash mines in western Canada. This led to a whole series of ever improving GPR measurements and work in this geological setting by the Geological Survey of Canada. These results were reported

by Annan et al (1988). Further coal mine developments were reported by Coon et al (1981).

In addition, the potential for use of borehole radar to investigate rock quality in potential hard rock nuclear waste disposal sites became a topic of interest. The Geological Survey of Canada and Atomic Energy of Canada supported thiswork (Davis and Annan (1986)).

Commercial instruments were used for most of this work and the number of activities spawned new commercialinterest. Geophysical Survey Systems Inc. remained the only supplier at this time but Ensco/Xadar was spawned in anattempt to create an alternate commercial product.

One major issue noted by the Geological Survey of Canada was the great difficulty in using existing equipment inremote areas. Equipment was heavy, bulky and power hungry. In addition, digital data was needed to exploit the dig-ital seismic processing advances rapidly evolving in the petroleum seismic field.

1980 – 1985

During this period, interest in GPR waned to a degree. The initial optimism for the technology gave way to the realitythat many environments were not favorable for GPR. Considerable confusion often existed as to whether failureswere equipment related or due to natural material responses. In addition, little money was available for technologydevelopment.

OYO Corporation of Japan developed a radar product called “Georadar” spawned by association with Xadar develop-ments. This instrument met some initial commercial success in Europe.

A-Cubed Inc. was formed in 1981 in Canada and started development of ground penetrating radars. The low fre-quency digital GPR developments were reported by Davis et al (1985). This technology development led to the pul-seEKKO series of GPR’s.

The nuclear waste disposal problem was continually studied and a number of countries funded the Swedish Geologi-cal Survey in the development of borehole radar. This work is reported by Olsson et al (1987).

Other applications for GPR such as road investigations and utility mapping met with mixed success. In general, thetechnology was quite new and not optimized for these applications. Work by Ulriksen (1982) provided a good foun-dation for some of these applications.

Many non-commercial developments occurred with prototypes that embodied the ideas for portability, digital record-ing and the use of fiber optics cables.

8/20/2019 GPR Annan 2003



6

Other little reported work was conducted by Southwest Research and the U.S. Army on borehole GPR to detect tun-neling in sensitive military areas (Owen (1981)).

1985 – 1990

GPR finally started to come into its own during this period. The strengths and weaknesses were becoming better understood and real problems in the near surface created a demand for high resolution mapping. The U.S. Environ-mental Protection Agency instituted many initiatives to investigate and clean up contaminated land (Benson et al(1984)). GPR was a natural tool to address high-resolution subsurface mapping and as a result a strong commercialdriver started to appear.

In addition, many of the previous applications were continually explored and movement to lower frequency GPR’swith full digital recording appeared in commercial products. Other applications such as soil classification for agricul-tural needs appeared (Doolittle & Asmussen (1992)). Adaption of one dimensional seismic modelling occurred in this period (Annan & Chua (1992)).

In 1988, Sensors & Software Inc. was spawned from A-Cubed Inc. and commenced commercialization of the pul-seEKKO technology.

1990-1995

The real explosion in the advancement of GPR occurred during this period. Many groups worldwide became inter-ested in the technology.

On the commercial side, Geophysical Survey Systems Inc. exhibited strong commercial success and was bought byOYO Corporation. During this period, Mala Geosciences was spawned from the Swedish Geological Survey roots.ERA in the UK also became more active using its research into unexploded ordinance and landmine detection to cre-ate commercial products. Sensors & Software Inc. grew rapidly broadening its pulseEKKO product line.

On the research side, much attention started to be paid by both the geophysical and electrical engineering community.Developments such as multi-fold data acquisition (Fisher et al (1992)), digital data processing (Maijala (1992), Ger-litz et al (1993)), and 2D numerical simulation (Zeng et al (1995), Cai and McMechan (1995)) occurred. Initial three-dimensional numerical simulation was reported by Roberts & Daniels (1996). Advances in applications in archeology

(Goodman (1994)), environmental (Brewster and Annan (1994)), geological stratigraphy using radar facies (Jol(1996)) and many other areas expanded. Environmental borehole GPR development was reported by Redman et al(1996).

Ground penetrating radar user meetings became more formalized and were held every 2 years at various locationsaround the world. This meeting provided a forum for the leading players in this field to meet, present results and dis-cuss problems. These meetings led to proceeding publications which are listed as references. These proceedings pro-vide a great deal of information for new users to the GPR field.

1995 – 2000

In this period, the evolution of the computers drove all of GPR advances. Numerical modelling of full 3D problems became more extensive albeit still with large computers (Holliger & Bergmann (2000), Lampe & Holliger (2000)).

The ability to manage the large volumes of information in digital form and manipulate them quickly became routine.As a result, acquisition of data on grids to make maps and grids and 3D visualization became practical (Grasmueck (1996), Annan et al (1997)). The commercial market and demand resulted in a variety of different and simpler sys-tems such as the Noggin from Sensors & Software Inc. (Figure 1-6).

8/20/2019 GPR Annan 2003



7

Figure: 1-6 Sensors & Software Inc.’s Noggin Smart Cart.

Strong research groups appeared at a number of universities. ETH led by Alan Green, the University of Texas at Dal-las led by George McMechan and the group at TU-Delft led by Jakob Fokema are some examples of groups pushingdevelopment of expertise and advancement of GPR frontiers.

The fundamental vector nature of GPR started to become critical. Understanding this full vector nature of the fields became of more interest (Roberts and Daniels (1996)). In addition, there was much more pressure on acquiring accu-rate positioning of information than historically needed because the need to do data manipulation requires very accu-rately controlled positioning information (Greaves et al (1996)).

1.3 CURRENT ACTIVITIES & FUTURE DEVELOPMENTS

GPR is now on very solid footing. Research groups with good understanding of the basic physics are developingmodelling tools and analysis capabilities. More work is still required on measurement of electrical properties of mate-rials. Electrical properties of mixtures are understood in general but the complexities and interactions in specificinstances are still subjects for research.

Digital processing power now exceeds our current capability to make use of it. As a result, development of softwareand processing algorithms to exploit the computer power available will cause an ever more rapid advance in themanipulation of data to address application needs.

Instrumentation is now stable and reliable. In the early days of GPR, instrumentation was always of marginal capabil-ity because of the extremely critical demands placed on the instruments. Designing ultra wideband antennas and elec-tronics to work in close proximity to a variable lossy dielectric media is not a trivial engineering exercise and onlynow are products becoming stable, reliable and reproducible.

Even now, the amplitude of GPR data is not well controlled. As instruments evolve and designs get better, the ampli-tude information of the data is becoming more reliable. Historically, the travel time was the most useful part of theGPR record. Relative amplitudes were good indicators but absolute amplitude information was unattainable.

As GPR becomes more sophisticated and stable, reliable quantitative amplitude information will spawn another gen-eration of data analysis and interpretation tools based on inversion to image material properties. Already inversion invarious forms to extract electrical properties is being attempted with success at the research level (van der Kruk (2001)).

A major trend already visible is the extraction of “user” information. In the early days of GPR, just acquiring the GPR

8/20/2019 GPR Annan 2003



8

cross section image in some form was the end goal of GPR measurement. With full digital data and analysis capabil-ity, display information in many forms is mundane and extraction of specific user information much more critical.

1.4 SUMMARY

The future of GPR is bright. The opportunities are still vast and new developments will occur at an every increasing

pace. Major points to note at present are as follows.

1. GPR is becoming a mature method.

2. Instrumentation is attaining a high level of quality and dependability.

3. Data processing and advanced presentation of imaging information is easy and is advanc-ing daily.

4. Instruments focused on specific applications and user needs are starting to appear and will be the way of the future.

The following sections are intended to show the basic principles underpinning the technology and the general state of practice at the current time.

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 2-Basic Electromagnetic Theory

9

2 BASIC ELECTROMAGNETIC THEORY

2.1 OVERVIEW

The foundations of GPR lie in electromagnetic theory. The history of this field spans more than two centuries and isthe subject of numerous treatise. The goal in this chapter is to provide the basic building blocks needed to work quan-titatively with GPR without reiterating reams of theoretical development. Unfortunately, a minimum set of tools must be established to make later discussions coherent and meaningful. Read the following with this in mind and be pre- pared to read reference material such as Jackson (1967), Reitz & Milford (1960), and Smythe (1989).

GPR represents a subset of the full electromagnetic field. GPR signals are electromagnetic waves; Maxwell's equa-tions which mathematically describe electromagnetic physics plus constitutive relationships which quantify material properties are the foundations for quantitatively describing GPR signals.

2.2 MAXWELL’S EQUATIONS

In mathematical terms, electromagnetic fields and related properties are expressed as:

(2-1)

(2-2)

(2-3)

(2-4)

Where:

- electric field strength vector

- magnetic flux density vector

- electric displacement vector

- magnetic field intensity

- electric charge density

- electric current density vector

Maxwell succinctly summarized the work of prior researchers in this compact form.

t

B

∂∂

−=Ε×∇

t

D J

∂∂

+=Η×∇

q D =•∇

0=Β•∇

E

B

D

H

q

J

8/20/2019 GPR Annan 2003


2-Basic Electromagnetic Theory GPR Principles, Procedures & Applications

10

A pictorial representation of the physics underlying of each equation are presented in Figure 2-1 to Figure 2-4. Equa-tion 2-1 summarizes Faraday's observation that a time varying magnetic field causes electric charges to move imply-ing the presence of a electric field.

Maxwell #1: Faraday Law

Figure: 2-1 A time varying magnetic field generates a closed loop electric field. A common example is the electric generator where a rotating magnet generates a voltage in a wire loop.

Maxwell #2: Ampere Law

Figure: 2-2 An electric current gives rise to a magnetic field. The classic example to demonstrate this concept isiron filings on a piece of paper forming circles about a current carrying wire.

8/20/2019 GPR Annan 2003



11

Maxwell #3

Figure: 2-3 Electric displacement starts (or ends) on an electric charge. Electric fields must form closed loops orterminate on a charge.

Maxwell #4

Figure: 2-4 Magnetic flux loops must close on themselves since there are no free magnetic charges.

D

8/20/2019 GPR Annan 2003



12

Ampere's observations that electric currents generate magnetic fields underpin Equation 2-2. Magnetic objects behave the same way when a magnet or an electric current is present. The electromagnet is the common example of this.

Equation 2-3 indicates that electric charges are sources (or sinks) of electric field. Electric fields emanate from elec-tric charges; hence Equation 2-3. Time varying electric fields will be of closed loop form when induction (Faraday'sobservation) occurs. The electric field will emanate outward (or into) when free charge is the field source. In general both field characters will be present and superimposed for time varying signals. Free magnetic charges have never been observed in nature; as a result, magnetic fields must form closed loops which explains Equation 2-4 and distin-guishes magnetic flux behavior from electrical field character.

From these building blocks, all classic electromagnetics (induction, radio waves, resistivity, circuit theory, etc.) can be derived after we characterize material electrical properties.

2.3 CONSTITUTIVE EQUATIONS

Constitutive relationships are the means of quantifying the physical properties of materials. In EM and GPR the elec-tric and magnetic properties are of importance. Constitutive equations provide a macroscopic (or average behavior)description of how electrons/atoms/molecules/ions etc., respond en masse to the application of a field.

For GPR, three quantities are defined, electric conductivity

(2-5)

which describes how free charges flow to form a current when an electric field is present. Dielectric permittivity,

(2-6)

which describes how constrained charges are displaced in response to an electric field.

Magnetic permeability,

(2-7)

which describes how intrinsic atomic and molecular magnetic moments respond to a magnetic field.

In general, , and are tensors and can also be non-linear. (i.e. ). For virtually all practical GPR issues,

these quantities are treated as field independent scalar qualities. (In other words, the response is in the same direction

as the exciting field and independent of field strength.) While these assumptions are seldom fully valid, the practicalworld of GPR has yet to be able to discern such complexity except in a few cases.

An additional feature of the properties is that they can depend on the history of the incident field. To be fully correct,

we should write Equation 2-5, Equation 2-6 and Equation 2-7 in the form (only Equation 2-5 is written for compact-

ness)

σ˜

Ε= σ ~ J

ε

Ε= ε ~ D

µ

Η=Β µ ~

σ ε µ σ σ E ( )=

8/20/2019 GPR Annan 2003



13

(2-8)

We will ignore this issue for now but will return to the topic later. (This more complex form of 2.8 must be used whenwe deal with frequency dependent or dispersive physical properties.)

For the rest of this chapter we will deal with scalar constant . Details on electric and magnetic properties will

be dealt with in later chapters. For GPR, and are of most importance in the majority of situations.

2.4 WAVE NATURE OF EM FIELDS

In this section the wave character of EM Fields is explored. Only the very simplest uniform medium is considered.Material properties are assumed isotropic, frequency independent and linear.

Maxwell's equations (Equation 2-1 through Equation 2-4) describe a coupled set of electric and magnetic fields whenthe fields vary with time. Changing electric fields create magnetic fields which in turn induce electric fields, as

depicted in Figure 2-5. This continuing succession of one field inducing the other results in fields which movethrough the medium. Depending on the relative magnitude of losses, the fields may diffuse or propagate as waves.With GPR we are most concerned with conditions where the response is wave-like.

Mathematically, the wave character is seen by rewriting Maxwell's equations to eliminate either the electric or mag-netic field. This is achieved by noting that the closed circulation character of the magnetic field can be expressed interms of the electric field. From Faraday's law (Equation 2-1) one can write:

(2-9)

Using Ampere's Law (Equation 2-2) plus the constitutive relatives one obtains:

(2-10)

which yields what is termed the transverse vector wave equation.

(2-11)

A B C

Terms B and C express the fact that currents generate magnetic fields whose time varying circulation create an elec-tric field as expressed by A. The equation represents that all fields must balance.

β )d β (t E )( σ ~(t) J o

−⋅β= ∫∞

ε µ σ, ,ε σ

( )Η×∇∂

∂−=Ε×∇×∇ µ

t

t t t

D J

t ∂Ε∂

−∂Ε∂

−=

∂∂

+∂∂

−=Ε×∇×∇2

µε µσ µ

∇ ∇ E µσ ∂ E

∂t ------ µε

∂2 E

∂t 2

---------⋅+⋅+×× 0=

8/20/2019 GPR Annan 2003



14

Figure: 2-5 Wave equation formulations express the self perpetuating field sequences as depicted.A moving chargeimplies current J which creates a magnetic field which induces an electric field which in turn causes electric charge

to move.

Note that we have eliminated the magnetic field in this formulation. If we had chosen to express our fields in terms of magnetic fields only, we would get exactly the same form of equation for H.

The essence of the transverse wave Equation 2-11 lies in it's basic solution form. The electric field (or magnetic field)vector must vary in a spatial direction perpendicular to the field vector.

Figure: 2-6 A field satisfying the transverse vector wave equation has the field vector perpendicular to the spatialvariation direction.

Mathematically, a position vector and a pair of orthogonal unit vectors ( ) and are needed to characterize a simplesolution to the EM wave equation. The electric field is

E

µ k ˆ,

8/20/2019 GPR Annan 2003



15

(2-12)

The key lies in understanding what the x operation implies. First we look at

(2-13)

Defining as the scalar distance in the direction that the field varies and we see

(2-14)

We see from Farady's law that the time varying magnetic flux

(2-15)

is created by the shearing electric field and is in a direction perpendicular to and namely (see Figure2-7) .

Continuing we find that

(2-16)

The shearing of the electric field generates a time varying orthogonal magnetic field which in turn creates an electricfield in the opposite direction to the original electric field.

Figure: 2-7 The electric magnetic and field variation direction form a 3D vector space

ut k r f ˆ,⋅=Ε

∇ ∇×

Ε×∇

β r k ⋅=

( )t f

uk ,ˆˆ β β ∂

∂×=Ε×∇

β ∂∂×−=Ε×∇−=∂Β∂ f uk t

ˆˆ

E k ˆ

w k ˆ

u×=

( )2

2

2

2

ˆˆˆˆ β β ∂

∂−=

∂∂

××=Ε×∇×∇ f

u f

uk k

8/20/2019 GPR Annan 2003



16

We find now that our fields in the simple case must have this vector character and the spatial/temporal variation satisfies the equation

(2-17)

Those versed in wave theory will recognize this as the scalar wave equation.

If we ignore losses (i.e. assume ), (Equation 2-17) reduces to

(2-18)

which has solutions of the form

( vt ) (2-19)

where

(2-20)

is the wave velocity. The wave nature is indicated by the fact that the spatial distribution of the fields translates in thedirection between observation times as depicted in Figure 2-8.

Figure: 2-8 A function of the form represents an event which moves spatially at a velocity v. Such functions

are solutions to the scalar wave equation.

β t ,( )

( ) ( ) ( ) 0,,,2

2

2

2

≡∂∂

−∂∂

−∂∂

t f t

t f t

t f β µε β µσ β β

σ o≡

( ) ( )

2

2

2

2 ,,

t

t f t f

∂∂

=∂

∂ β µε β

β

β t ,( ) f = β ±

εµ

1v =

β

β vt ±( )

8/20/2019 GPR Annan 2003



17

2.5 LOW LOSS – DAMPED EM WAVE CONDITIONS

For GPR we are most interested in problems where EM fields propagate as waves. All real materials generally exhibitsome loss which is primarily attributable to electrical conductivity. We will discuss this subject in more depth in later chapters.

In the previous section we tacitly assumed for convenience and saw how wave nature of EM fields flowed from

Maxwell's equations. The effect of small losses is of great interest in the real GPR world and we deal with this as a perturbation of the derivation of the previous section.

Returning to Equation 2-11, we can get a relative measure of the importance of various contributions to the waveequation by dimensional analysis. For a given spatial scale and time scale , the relative importance of terms inthe equation are given by the ratios

(2-21)

(2-22)

Our criteria for a low loss environment requires

(2-23)

or regrouping implies that

(2-24)

Electric currents in the material can be viewed as analogous to electric current in a resistor capacity circuit, asdepicted in Figure 2-9 (resistance is proportional to 1/ and capacitance is proportional to . The charging time for

the capacitor is expressed as / . The condition that |B/C| be small requires the time for charge redistribution to beshort compared to the time rate of change of the EM fields.

Figure: 2-9 A simple circuit containing a resistor and capacitor is the circuit analog to Equation 2-24. The resis-tance is proportional to 1/ σ and the capacitance is proportional to ε .

σ 0≡

x∆ t ∆

t

x

A

B

∆

∆≈

2

µσ

2

2

t

x

A

C

∆

∆

≈ µε

A

C

A

B⟨⟨

1⟨⟨

∆

= ε

σ t

C

B

σ εε σ

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



18

The solution of Equation 2-11 when reduced to the scalar form of Equation 2-17 is approximated by perturbing our wave solution when by defining

(2-25)

where p is a solution in the zero loss case, namely

(2-26)

Substituting into Equation 2-11 yields the result that

(2-27)

which we can satisfy if

(2-28)

and

(2-29)

In point of fact, we cannot satisfy the second condition but we can show that in the low loss situation this term is neg-ligible.

The first relationship is satisfied if

(2-30)

where

(2-31)

If we use Equation 2-30 as the solution for g , then

σ o≡

( ) ( ) ( ) β β β g ut pt f ±=,

2

2

2

2

t

p p

∂∂

=∂∂

µε β

02

22

2

=

±

∂

∂

∂

∂+

∂

∂ g

vu g p g p

σ

β β β

02

=±∂∂

g vu g σ

β

02

2

=

∂

∂

β

g

( ) αβ β ±= e g

2

vµσ α =

8/20/2019 GPR Annan 2003



19

(2-32)

Looking at the magnitude of the term

(2-33)

in Equation 2-27 and comparing it to C in our original Equation 2-11, we find the relative magnitude of this term to be

(2-34)

which is indeed negligible compared with all other terms in our original equation when conduction losses are small.

Our solutions are damped waves of the form

(2-35)

which decay exponentially in amplitude in the direction the wave is travelling. α is called the attenuation coefficient.The EM field translates with shape invariance as a wave but decreases in amplitude as energy is lost to ohmic dissipa-tion in the material. Figure 2-10 illustrates the concept of how the field changes with time and distance.

Figure: 2-10 EM fields propagate as spatially damped waves when electrical losses are small. The signal amplitudedecays exponentially in the direction of field translation while the field shape remains invariant.

g v g

2

2

2

2

=∂

∂ µσ

β

pg v g

p2

2

2

2

=

∂∂ µσ

β

1

4

12~

2

2

2

2

⟨⟨

∆=

∆

∂∂

ε

σ

µε

µσ

β t

t

v

c

g p

( ) ( ) αβ µ β β ±±= et pt f ,

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



20

2.6 SINUSOIDALLY TIME VARYING FIELDS

Many EM developments by-pass the discussions given in the previous two sections and deal with EM fields assumingsinusoidal time variation. This approach is in essence the use of a Fourier transform approach to separate time andspatial operators.

In simple terms, the form of f is assumed to be

(2-36)

and the partial differential Equation 2-11 reduces to the equation

(2-37)

Defining

(2-38)

the solution for h has the form

(2-39)

k is called the propagation constant. Quite often it is convenient to express k in the form

(2-40)

where v is the phase velocity (see 2-20) and a is the attenuation. When this is done we have f in the form

(2-41)

which is exactly the same form of translating and attenuated field behaviour noted in the previous two sections.

The Fourier (or Laplace) transform solutions are necessary in generalized formulations since the sinusoidal excitationsolutions are exact. To generate time domain transient solutions the inverse Fourier or Laplace transformation isapplied to the frequency domain solution. Often this must be done numerically as analytical solutions are seldom possible.

When losses are low, we find that the two terms in the propagation constant, k, are such that

(2-42)

( ) ( ) t ieht f ω β β =,

d 2h

d β2

--------- ω2εµ iωεµσ+( )h – 0=

k ω2εµ iωµσ+( )

1

2---

=

( ) β β ik eh ±Α=

α ik +=v

( )

±

±Α= vt i

eet f

β ω

αβ β ,

ωµσ ω2εµ«

8/20/2019 GPR Annan 2003



21

or

(2-43)

which is called the low loss criteria.

When this condition holds, we find

v = (2-44)

(2-45)

which is the same as our analysis in the previous sections.

2.7 ELECTROMAGNETIC IMPEDANCE

In Section 2.4, the basic solutions of Maxwell's equations were seen to be coupled electric and magnetic fields. Thefields are perpendicular to one another and move in a direction perpendicular to both of the fields as depicted in Fig-ure 2-7. As indicated earlier, both the electric and magnetic fields satisfy the same transverse wave equation. Solvingfor one field allows an immediate solution for the other.

When these prior simple solutions were developed, we worked in terms of the electric field. The amplitude of the

magnetic field can be directly related to the electric field and vice versa because of the field coupling. A commonterm in electrical engineering is the electromagnetic impedance defined as

(2-46)

From section 2.4,

(2-47)

and

(2-48)

since

1⟨⟨ωε

σ

1

εµ------

1

2---

2

vσ α =

ΗΕ

=Ζ

( )ut f ˆ, β =Ε

µ

Β=Η

8/20/2019 GPR Annan 2003



22

(2-49)

we have

(2-50)

making

(2-51)

Substituting in we find

(2-52)

which states that the relative amplitude of electric and magnetic fields in our solutions (known as plane waves) aredetermined by the medium properties.

When conductivity is important, we must deal with sinusoidal signals to make math simple and we find that

(2-53)

In some cases, the inverse of Z, called the electromagnetic admittance, is used and is expressed as

(2-54)

These concepts are very useful and will take on significance in later chapters.

In low loss environments, the approximate forms are

t

f k u f k u

t ∂∂×

=∂∂

×=∂∂

v

ˆˆˆˆB

β

]v

[ 0 f

u k ˆ Bt

=×−∂∂

v

f

u k ˆ

B ×=

εµ

=µ=Ζ v

ω

σ ε

µ

i+=Ζ

µ

ω

σ ε i+

=Ζ

=Υ1

8/20/2019 GPR Annan 2003



23

(2-55)

2.8 POLARIZATION

From the intrinsic vector nature of the electromagnetic fields studied, we see the electric and magnetic field vectorsare orthogonal to each other and to the direction of translation. From this result, two independent electromagneticfields can exist for a given propagation direction. Figure 2-11 depicts the concept.

Figure: 2-11 For a given propagation direction , two independent electromagnetic fields, a and b, exist.

By convention, the EM field solutions are characterized by the direction of the electric field vector. Plotting the elec-

tric field vector in a plane perpendicular to one observes

Figure: 2-12 The total observed field, is the vector sum of the two independent fields, and . The polariza-

tion direction is in the direction of .

When the time variation of the fields is sinusoidal, the concepts of linear, circular and elliptical polarization arise.The electrical field vector has the form

ωε

σ+

µ

ε≈

ωεσ

−εµ

≈Ζ

2i1Y

2i1

k

k ˆ

E E a E b+=

E E E b2

+( )1 2 ⁄

=

θtan E b E a-----=

E E a E b

E

8/20/2019 GPR Annan 2003



24

(2-56)

where Ea and E b are the scalar amplitudes, and φa and φ b are the phase of each of the components with respect to acommon reference and and are unit vectors perpendicular to one another.

If , the electric field is said to be linearly polarized. The vector is fixed in direction (i.e. θ = constant) and itsamplitude varies sinusoidally as depicted in Figure 2-13.

(2-57)

Figure: 2-13 When fields vary sinusoidally with time, linear polarization manifests itself as a total electric fieldwhich is in a fixed direction.

beaee t iei

b

t ii

aba ˆˆ ω φ ω φ Ε+Ε=Ε

a b

φa φb=

t ba cos)( 2

122

Ε

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



25

Figure: 2-14 For sinusoidal time varying fields, the fields are, in general, elliptical. Special cases are linear andcircular polarization.

In the elliptically polarized case, the amplitude and field direction change with time tracing out an ellipse in the field plane and an elliptical spiral in space, as depicted in Figure 2-14. More discussion of polarized fields can be found inany of the EM texts referenced at the end of the chapter.

Linear, elliptical and circular polarization have limited meaning when dealing with transient fields. In such cases, thefield vector may rotate regularly for short durations but generally motion is irregular. Plots of field diagrams showingamplitude and direction for transient events versus time (Figure 2-15) called hodograms - can be very unstructured.

8/20/2019 GPR Annan 2003



26

Figure: 2-15 Plotting the total transient electric field vector in a plane versus time creates a hodogram. The arrowtraces the vector as a function of increasing time.

2.9 EM FIELDS FROM SOURCES

To this point, our analysis has carefully avoided dealing with sources. This subject becomes mathematically complex. In the following a very cursory view of the subject is given with the objective of establishing fundamental con-cepts needed for later chapters.

For most GPR applications, we excite electromagnetic fields by causing electrical currents to flow on metal structures(called antennas). In a similar fashion we detect electromagnetic fields by the currents they induce to flow on similar metal structures.

For simplicity, we assume we can separate our electric current in Ampere's law (Equation 2-2) into two parts.

(2-58)

where is our source or exciting current and is the conduction current as defined in 2-5. How is created in

a controlled fashion is not a trivial issue and one we will skip over for now.

Following on, we find our wave Equation 2-11 is modified to the form

(2-59)

The solution to this equation often requires resorting to numerical calculations. There are some simple instances of analytical solution and one of these we will use for considerable discussion throughout.

The solution for a small current element can be derived in analytical form in a uniform medium. We treat the currentdensity as being concentrated (Figure 2-16) in a line of infinitesimal diameter and length.

Ε+= σ s J J

J s σ E J s

t

J

t

E

t

EE s

2

2

∂

∂µ−=

∂

∂µε+

µ

∂µσ+×∇×∇

8/20/2019 GPR Annan 2003



27

Mathematically we describe

(2-60)

where δ(x) is the Dirac delta function.

Figure: 2-16 A small electric dipole antenna can be though of as a small current element.

Aside: Physically we treat to exist in a small parallel piped as shown in Figure 2-17.

Figure: 2-17 Depiction of a parallel piped of current mode by using boxcar function B( µ ,b).

where

2ˆ)()()( e z y x I J δ δ δ l∆=

J s

( ) ( ) ( ) 2ˆ,,, e z z y x x

z x

I J s ∆Β∆Β∆Β

∆∆= l

8/20/2019 GPR Annan 2003



28

We let ∆l, ∆ x, ∆ z approach 0 and in the limiting case

subject to the condition remaining finite.

This mathematical trick of taking of the current to an infinitesimal size permits us to ignore details of internal sourcegeometry. This source is called an infinitesmal electric dipole.

Without going into the detailed proof, the solution for field generated for this infinitesimal current source takes the

form

(2-61)

where

(2-62)

The kernel, g , is somewhat like our previous solutions to the sourceless case in that signals are shifted (delayed) intime by distance traveled but in this case the signals emanate spherically outward from our source point and decrease

with distance. For more extensive discussion see the EM texts on the subject of retarded potentials and retarded fieldsolutions.

As with our previous discussion, when the electrical losses are small, (µσ small but finite), our expression is modifiedto read.

( )

otherwise 0

2 to

2 1,

=

>−>=Βbb

b µ µ µ

( ) ( ) ( )

( ) ( ) ( )

0z ,0 ,0

ˆ

ˆ,,,

2

2

→∆→∆→∆

∆→

∆

∆Β

∆

∆Β

∆

∆Β∆=

l

l

l

ll

x

e z y x I J

e z

z z y

x

x x I J

s

s

δ δ δ

I ∆l

E

∂∂

∇+=Ε y

g g e2

ˆ

2

1222 )(

4

)v(),,,(

z y xr

r

t r I t t z y x g

++=

−∂∂

∆=

π

µ l

8/20/2019 GPR Annan 2003



29

(2-63)

Frequently one finds solutions to the finite source problem derived for sinusoidally time varying fields as we have

discussed earlier. When the fields are sinusoidal in time, Equation 2-59 becomes

(2-64)

and the solutions kernel g takes the 2-63 form

(2-65)

The time lag becomes a phase shift dependent on distance from the source and the time derivative of the current becomes a multiplication by frequency.

2.10 FAR FIELD APPROXIMATION

When the observer is far from the source, the solutions simplify considerably. Descriptions at a distance are referredto as "far-field approximate" solutions. For (or )the field has a general form

(2-66)

where and depends on the angular position of the observer with respect to the source depicted in

Figure 2-18.

Figure: 2-18 The electric and magnetic field at large distance from the source described using a spherical polarangle notation. E and H become perpendicular to the radial direction from source to observer.

r

et r I t t z y x g

ar

π

µ

4

)v(

),,,(

−−∂∂

∆=

l

s J ii ωµ εµ ω ωµσ =Ε−Ε−Ε×∇×∇ 2

r

ee I i z y x g

ar

r i

π

ωµ ω

4),,(

vl∆=

E H

( ) ( ) g v φ φ θ θ φ θ µ ˆ,ˆ, +=Ε

µ θ φ,( ) v θ φ,( )

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



30

In the far field, and are orthogonal (perpendicular to one another) and have no radial component. Whenviewed over a limited spatial volume, the field resembles a "plane wave" discussed earlier. The electric and magneticfields and the radial vector from the source are mutually orthogonal as depicted in Figure 2-19.

Figure: 2-19 At a large distance from the source, the fields propagate radically outward and resemble plane waves.

In fact,

(2-67)

where Z is the electromagnetic impedance as discussed in section 2.9. is the unit vector in the radial direction.

2.11 ANTENNA PATTERNS, DIRECTIVITY, GAIN

All these terms derive from the far field approximation. In physical terms, the far field represents the energy lost fromthe source as radiated electromagnetic energy. The best way to view far field radiation is to visualize measuring thefield at the surface of a large sphere which encloses the source. In theoretical terms the far field approximation becomes exact as the radius of the sphere becomes infinite.

Position on the surface of the sphere is defined by polar (angle) coordinates ( ) as depicted in Figure 2-18 and2.20. (i.e., like latitude and longitude but latitude angles are measured from the north pole rather than the equator).

At any point on the sphere the electric (and magnetic) field is tangential to the surface. The field vector is usually

decomposed into components which are in the and unit vector directions. and are tangential to the sphere

surface and perpendicular to one another. The field vector is then expressed as

(2-68)

The same applies to .

Ε Η

( )Η×Ζ=Ε r

r

θ φ,

θ φ θ φ

φ θ φ θ ˆˆ Ε+Ε=Ε

Η

8/20/2019 GPR Annan 2003



31

As discussed earlier, the fields are polarized and the direction of is normally used as the polarization direction.

Figure: 2-20 Far field antenna patterns are the fields on a very large sphere. Position on the surface is measuredin spherical polar angles as indicated The power radiated by the source is measured by the energy in the electromag-

netic field transported across the sphere's surface.

The power radiated by the source is measured by the energy in the electromagnetic field transported across the

sphere’s surface. A vector called the Poynting vector is a measure of power crossing the sphere at any point.

(2-69)

is directed radially outward at all points on the sphere. The total radiated power is expressed as:

(2-70)

where integration is over the surface area of the sphere.

After this preamble we can get down to some specifics. Sources of electromagnetic radiation seldom emit energyuniformly in all directions. The intensity of the far field varies with position ( ). As the cartoon in Figure 2-21shows, a highly direction source only emits energy in a small cone.

Ε

P

Ζ

Ε=Η×Ε=Ρ

2

P

APPT

d r ˆ •= ∫∫

θ φ,

8/20/2019 GPR Annan 2003



32

Figure: 2-21 Conceptual illustration of a highly direction source. The radiated energy results in an electric field,Eo , only in the small shaded area of the far field sphere surface, elsewhere the electric field is zero.

Directivity can be expressed for a specific field component or for total field.

(2-71)

(2-72)

(2-73)

Generally ERef. is selected so the maximum directivity is 1 and is therefore application and source type dependent. For example, some sources can result in so that

(2-74)

A related concept is antenna gain. Gain deals with power rather than field strengths. As with directivity it may becomponent specific. More often gain is related to total power rather than component power. Gain is a most useful parameter in system engineering.

Returning to the Poynting vector discussion, the total energy emitted by a source is

f ReE

),(E),(D

φθ=φθ θ

θ

f ReE

),(ED

φθ= φ

θ

.f Re

2

122

tE

)),(E),(E(),(D

φθ+φθ=φθ φθ

E φ 0≡

),(D),(Dt

φθ≡φθ θ

8/20/2019 GPR Annan 2003



33

(2-75)

If the source radiated uniformly in all directions (called an isotropic radiator source), then

(2-76)

would be constant. The fact that P varies in most cases indicates that more energy is sent in one direction thananother. In other words, some directions are favored relative to others. Gain (or loss) is a relative measure. Antennagain is defined as

(2-77)

which is the ratio of the local power crossing the sphere relative to the power which would be observed if the samesource emitted the same power uniformly in all directions (i.e. isotropically).

In practice, G is often expressed in decibels.

(2-78)

Gain, as indicated earlier, is a relative measure of signal enhancement or depletion.

Both directivity and gain have historically been displayed as functions on a cross section of the sphere since plottingdata on a spherical surface (translated to a flat page) is difficult.

The antenna pattern is created by drawing a point at a radial distance from the great circle origin proportional to G,GdB or D. The concept is illustrated in Figure 2-22, Figure 2-23 and Figure 2-24.

Figure: 2-22 A cut through the center of the sphere is used to generate a circular flat plane for display of signalvariations in that cross section of sphere.

∫∫

∫∫

•φθ=

•×=

Aˆ),(P

Aˆ)HE(PT

d r

d r

2

T

4

P),(P

r π=φθ

2

T

4

P

),(P),(G

r π

φθ=φθ

θ φ,( )

),(Glog10),(G10dB

φθ=φθ

8/20/2019 GPR Annan 2003



34

Figure: 2-23 The antenna pattern is created by a vector with length proportional to gain or directivity and in the given direction.

The line that the end of the vector traces as vary defines the pattern.

Figure: 2-24 An example of antenna gain pattern.

Antenna patterns (either gain or directivity) are displayed in this fashion. More specific examples will appear later inthese notes.

θ φ,( )

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 3-Physical Properties I

35

3 PHYSICAL PROPERTIES I

3.1 WHY ARE PHYSICAL PROPERTIES IMPORTANT?

GPR investigates the subsurface by making use of electromagnetic fields which propagate into the subsurface. EM

fields which are time varying consist of coupled electric (E) and magnetic (H) fields. As discussed in section 2 thefields interact with the surrounding media. This interaction is macroscopically described by the constitutive equa-tions 2.5 to 2.7. The manner in which the electromagnetic fields interact with natural materials controls how electro-magnetic fields spread into the medium and are attenuated in the medium. In addition, the variation in physical properties gives rise to the observed subsurface reflections obtained with a GPR system.

In most geological and NDT (non-destructive testing) applications of GPR, electrical properties tend to be the domi-nant factor controlling GPR responses. Magnetic variations are usually weak. Occasionally magnetic properties canaffect radar responses and GPR users should be cognizant of magnetic effects.

An electric field in a material gives rise to the movement of electric charge, (i.e., electric current). The current flowdepends on the nature of the material. There are two types of charge in a material, namely bound and free, which giverise to two types of current flow, namely displacement and conduction. In the following, we will provide a simpleoverview of the two types of current flow. An in-depth discussion of electrical properties can be found in the text byVon Hippel, (1954).

Magnetic properties are controlled by the electric charge circulation character at the atomic and molecular level.Macroscopic magnetic properties are addressed briefly in these notes. Von Hippel (1954) addresses some of the basicconcepts.

3.2 CONDUCTION CURRENTS

Most people are very familiar with electrical conduction currents. Conduction currents are created when unbound(free) charges move in a material. The electrons which flow in a metal wire are an example of conduction current. Ina metal, electrons move through the metallic matrix to transfer charge from one point to another. Another commonconduction mechanism is the movement of ions in a water solution. The later is much more important in most GPR applications.

Conduction currents arise when free charge accelerates to a terminal velocity (basically instantaneously) when anelectric field (E) is applied. As long as the electric field is applied, the charge moves; when the electric field isremoved, the charge decelerates and stops moving Figure 3-1 illustrates these concepts.

Figure: 3-1 Conceptual illustration of charge movement for conduction currents.

8/20/2019 GPR Annan 2003


3-Physical Properties I GPR Principles, Procedures & Applications

36

a) Charge velocity versus time after E field applied.

b) Energy is extracted from the applied electric field versus time.

Figure: 3-2 When an electric field is applied, unbound electrical charges accelerate to a terminal velocity. Afterinitial acceleration, velocity becomes constant and a continual transfer of energy to the surrounding material in the

form of heat occurs

All the time that charge is moving, the moving charge is working against its surroundings dissipating energy in theform of heat. The moving charge bumps into 'non-moving' objects and transfers mechanical energy which appears inthe form of heat in the medium. Conduction currents represent an energy dissipating mechanism for an electromag-netic field. Energy is extracted from the electromagnetic field and transferred irreversibly into the medium as heat.

Mathematically one describes the relationship between conduction current and the applied electric field as indicatedin Equation 3-1.

(3-1)

In simple materials, the relationship is linear and the proportionality constant is referred to as the electrical conductiv-ity. Electrical conductivity has units of Siemens per meter (S/m). For many applications, however, it is more usefulto work with units of milliSiemens per meter (mS/m). Conductivity is dependent on the charge density and the inter-nal statistical mechanical interaction of the charge with its surroundings. These details are beyond the scope of thisdiscussion.

Ε= σ ~ J

8/20/2019 GPR Annan 2003



37

It should be noted that electrical conductivity and resistivity are directly related. Refer to Figure 3-3 for the relation-ship and the expression of Ohm's law. Electrical resistivity is the inverse of electrical conductivity.

Figure: 3-3 Relationship between current and applied field as well as the relationship will Ohm's law and resis-tively.

It is important to note that there are simplifications in the above discussion from the general form shown in Chapter 2.The conductivity is shown as being a constant. In fact it can be a function of the rate of change of the electric field,the amplitude of electric field itself, as well as temperature, pressure and many other factors. As a result, one shouldnot be surprised to see both non-linearity and frequency dependent conductivity in real materials. Generally these aresecond order effects but they must be considered when advanced use of GPR is contemplated. For this basic GPR overview, they will be treated as secondary issues.

3.3 DISPLACEMENT (POLARIZATION) CURRENTS

Displacement currents are associated with bound charges which are constrained to limited distance of movement.Examples of this are the electron cloud around an atomic nucleus, the electrical charge in a small metal object imbed-ded in an insulating environment, and the redistribution of the molecular dipole moment intrinsic to some molecules.Figure 3-4 depicts the concept. When an electric field is applied, bound charge moves to another static configuration.This transition occurs virtually instantaneously after which the charges no longer move. During the transition, energyis extracted from the electric field and the energy is stored in the material. When the field is removed, the chargemoves back to the original equilibrium distribution and energy is released. This type of behavior is typical of whathappens in a capacitor in an electric circuit. Energy is stored by the build up of charge in the capacitor and thenenergy is extracted by the release of that charge from the device.

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



38

Figure: 3-4 Conceptual illustration of charge movement associated with displacement currents.

Figure 3-5 depicts the characterization of charge separation in a material. When an electrical field is applied, displacement of charge in a bulk material gives rise to a dipole moment distribution in the material. The charge separa-tion is described in terms of a dipole moment density, D. In a more formal derivation, D is called the electricdisplacement field (see chapter 2). In simple materials, the induced dipole moment density is directly proportional to

the applied electric field and the proportionality constant is referred to as the dielectric permittivity of the materialand has units of Farads/m (F/m).

Figure: 3-5 Dipole moment density induced by applied electric field and relation to displacement current.

The creation of a dipole moment distribution in the material is associated with charge movement. The electric currentassociated with this charge movement is referred to as displacement current. The displacement current is mathemati-cally defined as the time rate of change of the dipole moment density.

The electric permittivity is never zero. Even in a vacuum, the permittivity, takes on a finite value of 8.85 x 10-12

F/m (Farads per meter). The explanation for this lies in the field of quantum electrodynamics and is far beyond thescope of this discussion.

It is often more convenient to deal with a dimensionless term called relative permittivity or dielectric constant, K. Asdepicted in Figure 3-6 , the relative permittivity is the ratio of material permittivity to the permittivity of a vacuum.

ε0

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



39

Figure: 3-6 Dielectric constant or relative permittivity is the ratio of permittivity of material to that of free space.

3.4 TOTAL CURRENT FLOW

In any natural material, the current which flows in response to the application of an electric field is a mixture of con-duction and displacement currents. Depending on the rate of change of the electric field, one or other of the two typesof current may dominate the response. Mathematically, the total current consists of two terms; one which depends onthe electric field itself and one which depends on the rate of change of the electric field.

(3-2)

(3-3)

Quite often it is useful to deal with sinusoidally time varying excitation fields. In this situation, one finds that the displacement currents are proportional to the angular frequency.

(3-4)

where is the angular frequency.

The displacement currents are out of phase with the conduction currents by 90° which is what ascribing an imaginary

aspect to the displacement component implies. Those familiar with electrical engineering circuitry termi-

nology will realize that there is a phase shift between the conduction currents and the displacement currents whichindi-

cates that one term is an energy dissipation mechanism and the other one is an energy storage mechanism .

DCJJJ +=

dt

dΕσΕJ ∈+=

Ε∈+= )(J σ i

ω

i 1 – =

8/20/2019 GPR Annan 2003



40

A simplified plot of displacement current and conduction currents as well as total current versus frequency is pre-sented in Figure 3-7. Usually there is some frequency above which the displacement currents exceed the conductioncurrents. In a simple material where the conductivity and the dielectric permittivity are constant, there is a transitionfrequency, , where the displacement currents and conduction currents are equal. Above this frequency, displace-ment currents dominate; below this frequency, conduction currents dominate. This factor is important when we dealwith EM wave propagation. This frequency defines the on set of the low-loss regime important to GPR.

Figure: 3-7 Conduction, displacement and total current versus frequency.

Mathematically the transition frequency is defined.

(3-5)

In addition, another term called the loss tangent is defined. The loss tangent is the ratio of conduction to displace-ment currents in a material.

(3-6)

The term loss target tends to be most common in electrical engineering contexts.

Conductivity and permittivity are not independent of the excitation frequency. There is always some variation. Thistopic is beyond the scope of this chapter of the notes but there is considerable literature (i.e., Olhoeft, 1975) on fre-quency dependent electrical properties which can be referred to.

3.5 MAGNETIC PERMEABILITY

Magnetic permeability is seldom of major importance for GPR applications. For completeness and to address thoseexceptional situations where permeability may become important, we review some of the basic aspects of magnetic permeability.

Magnetic permeability is actually related to the intrinsic electrical characteristics of the basic building blocks of phys-ical materials. In simple terms, charged particles, which form atoms, which in turn make molecules, have a quantummechanical property referred to as spin. When combined with the charge on the particle, spin results in the particlehaving a magnetic dipole moment. When an electron moves around an atomic nucleus, the charge motion can alsocreate a magnetic moment.

t

∈

σ

tω

∈ω

σ

J

J

δtanD

C

8/20/2019 GPR Annan 2003



41

The simple analogy is to have electrical charge uniformly distributed on a spherical ball and then spin ball. Theresulting rotating charge appears to be a circular loop of current, which in turn gives on rise to a magnetic dipole.Magnetic properties are essentially the properties of an electrical change moving around a closed path.

a)

b)

Figure: 3-8 a) A simple picture suggesting the origin of electron spin movement; b) relating the magnetic momentinduced in an electron cloud by a change in magnetic field.

8/20/2019 GPR Annan 2003



42

Figure: 3-9 Relating the magnetic moment to a simple electron orbit.

The details are obviously more complex but this provides a simple pictorial model to use. Atoms are formed of elec-trons and protons plus neutrons. The electrically changed components have intrinsic and orbital spin when they formmolecules of a given type of material. The particular orientation of the spin axes of the individual particles can bealigned in random or structured ways and may be altered by an ambient magnetic field. If the molecular structuredoes not accept random spin orientation but requires a structured crystalline architecture, the material can have a per-manent magnetization. If component parts can move to align parallel or anti-parallel to an applied field, an inducedmagnetization response will arise.

Magnetic permeability measures the degree to which individual dipole moments of the building blocks can be alignedor moved from their normal orientation by an externally applied magnetic field. The more of the individual momentsthat can be moved into alignment, the more magnetically polarizable the material. The magnetic properties of materi-als are quantified by magnetic dipole moment density. When an electrical current flows in a closed loop, the mag-netic moments is

(3-7)

where M is the dipole moment, I is the current and A is the area of the loop enclosed by the current filament. M has

units of Am2. For bulk materials, the material is characterized by dipole moment density

(3-8)

which has units of A/m. V is volume. When a magnetic field, H, induces a magnetic moment, the amount of momentis expressed as

(3-9)

nΙΑ=Μ

V m

Μ

=

Η= k m

8/20/2019 GPR Annan 2003



43

where k is the magnetic susceptibility (and is a dimensionless quantity). There is considerable similarity betweeninduced magnetic moment and induced electric dipole moment discussed previously in the displacement current sec-tion.

In the material, the magnetic flux is expressed as

(3-10)

and magnetic permeability is expressed as

(3-11)

where

H/m (3-12)

The term relative magnetic permeability is expressed as

(3-13)

in an analogous fashion to relative permittivity. When both magnetic and electric properties vary, relative permittiv-ity is usually expressed as K e to avoid confusion.

The presence of a magnetic field induces the individual dipole moment to change orientation and line up with theapplied field. In some materials the alignment is in the same direction as the applied field, whereas in other materialsthe alignment maybe anti-parallel to the applied field. These two types of behavior referred to as paramagnetism anddiamagnetism. Generally these responses are very weak and give rise to small variations in magnetic permeability.

Typical values of magnetic susceptibility are less than 10-5.

In some situations, however, the magnetic moments can be aligned in large sections (called domains) of the crystalstructure of a material. The moment of domains can be changed by the molecules in the crystal structure behaving ina sympathetic fashion and moving from one domain to another. Such materials are known as ferromagnetic materi-als.

In ferromagnetic materials, the polarization can be quite large and high values of K m in the range of tens or even hun-dreds may be observed in materials such as iron, cobalt, and nickel. With ferromagnetic materials, the behavior ismore complex in that the dipole moments when moved, or aligned, remain aligned. This is known as permanentmagnetization. In such materials the permeability is very high and the dynamic behavior of the material complex.Such materials seldom form a large volume fraction of soils and rocks but their presence in small amounts can be thedominant factor determining bulk permeability.

Behavior can be very complex. The behavior of dipole moment density is controlled by how domains move, growand change orientation which can be field dependent, frequency dependent and temperature dependent. The subjects

are well beyond the scope of these notes.

( )m H B += 0µ

( )k += 10µ µ

µ0 4π 107 –

×=

( )k K m +== 10µ

µ

8/20/2019 GPR Annan 2003



44

Figure: 3-10 Ferromagnetic domain structures: a) single crystal, b) polycrystalline specimen. Arrows representthe direction of magnetization.

Figure: 3-11 Magnetization of a ferromatic material: a) unmagnetized, b) magnetization by domain wall motion, c)magnetization by domain rotation.

8/20/2019 GPR Annan 2003



45

In soils and rocks, magnetic behavior is dictated by the amount of magnetite (or similar minerals such as meghemiteor ilmenite). The graph (from Grant & West (1965)) in Figure 3-14 shows how susceptibility varies with magnetitevolume fraction. The simple approximate formula is

(3-14)

where q is the volume fraction of magnetite in the material. To put this result in perspective, 1% by volume magnetite(which is very high in most cases) content yields K m=1.038. Only in rare cases will K m be significantly different fromunity.

Figure: 3-12 Data from which the emperial formula for susceptibility k=2.89 x 10-3 V1.01 was derived.[Mooney and Bleifuss (7).]

3.6 COMPLEX MATERIALS

GPR measurements are seldom carried out in pure elements. Almost all of GPR work is carried out where materialswhich are composites of many other materials or elements. GPR surveys on water represent one of the few applica-tions on a "pure" single element material.

Table 3-1, Table 3-2 and Table 3-3 show in simple form the composition of some material, which might be encoun-tered in GPR applications. Table 3-1 is a simple quartz sand. This material contains a mixture of soil grains, air,water and ions dissolved in water in variable proportions. Normally soils grains occupy 60 to 80% of the volume.

θ 8.3=k

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



46

Table: 3-1 Components of Quartz Soil

quartz grainsair

waterdissolved ions

Table 3-2 summarizes the constituents of hydrocarbon contaminated simple quartz sand such as described in Table 3-1. In addition hydrocarbons in both liquid and vapor form plus a variety of different hydrocarbon compositions will be present as well as biodegraded hydrocarbon derivatives.

Table: 3-2 Components of Hydrocarbon contaminated Quartz& Soil

quartz grainsair

dissolved ionswater

liquid hydrocarbonbio degradation products

The third example given in Table 3-3 is concrete consisting of cement and aggregate. The aggregate can be any mix

of stone materials which are appropriate for mixing with cement and of variable mineralogy. The cement can be fullhydrated or partially hydrated. An optimal mix will have very little pore space containing water and/or air. A lessthan optimal mix will have substantial pore space, which may be connected or disconnected, and which can containair and/or water.

Table: 3-3 Components of Concrete

aggregatecement hydrated

cement partially hydratedair

waterdissolved ions

These are but three simple examples and many more could be cited. Understanding GPR responses and their rela-tionship to electromagnetic properties thus becomes the subject of understanding the properties of mixtures of mate-rials. Mixtures of materials lead to physical property variation which are seldom in proportion to the volume fractionof the constitute components electromagnetic responses. The subject is not simple and is an area of current research.In many respects, this complexity can make quantitative analysis of GPR data impossible without ancillary informa-tion.

For example, soil density will have some impact on the GPR response because it will ultimately determine pore spaceavailable. Unfortunately density only weakly affects the bulk permittivity whereas the presence of water (the amountof which is controlled by porosity which reflects density) has a large effect on both the permittivity and the conduc-tivity. Using GPR observations to infer soil density requires addition constraints such as full water saturation (i.e.material below the water table).

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



47

Table: 3-4 Typical Dielectric Constant, Electrical Conductivity, Velocity and Attenuation

Observed in Common Geologic Materials

While the subject of mixtures is quite complex, the big picture is not as bad as it sounds. Table 3-4 summarizes therelative permittivity and conductivity for some common materials encountered with GPR (typically in the 10 to 1000MHz frequency range) . The feature that stands out is the presence or absence of water in the material.

A very general picture emerges which the following points summarize.

• Bulk minerals and aggregates in mixtures generally are good dielectric insulators. They typi-

cally have a permittivity in the range of 3 to 8 and are usually insulating with virtually zero

conductivity.

• Soils, rocks, and construction materials such as concrete, asphalt, etc. have these constituent

elements with empty space between the grains (pore space) available to be filled with air, water

or other material.

• Water is by far the most polarizable material (in other words it has a high permittivity) natu-

rally occurring.

• Water is invariably present in the pore space of natural (Earth) materials except such unique

situations where vacuum drying or some other mechanism assures the total absence of water.

• Water in the pore space normally contains ions and the water conductivity associated with ion

mobility is the dominant factor in determining bulk material conductivity.

MATERIAL K (mS/m)

v

(m/ns)

a

(dB/m)

Air 1 0 0.30 0

Distilled Water 80 0.01 0.033 2xl0-3

Fresh Water 80 0.5 0.033 0.1

Sea Water 80 3xl03 .01 103

Dry Sand 3-5 0.01 0.15 0.01

Saturated Sand 20-30 0.1-1.0 0.06 0.03-0.3

Limestone 4-8 0.5-2 0.12 0.4-1

Shales 5-15 1-100 0.09 1-100

Silts 5-30 1-100 0.07 1-100

Clays 5-40 2-1000 0.06 1-300

Granite 4-6 0.01-1 0.13 0.01-1

Dry Salt 5-6 0.01-1 0.13 0.01-1

Ice 3-4 0.01 0.16 0.01

σ

8/20/2019 GPR Annan 2003



48

Empirically derived relationships shown in Figure 3-13 and Figure 3-14 demonstrate the relationship between the permittivity and conductivity of materials with variable volumetric water content. As a general rule the permittivityat zero volumetric water content is in the 3 - 4 range and conductivity is usually very small. As water is added to themix, the permittivity and conductivity rise until no more water can be squeezed into the available space in the mix-ture. Obviously porosity of the material dictates the maximum limit on the volume of water that can be placed in thematerial and ultimately that in turn dictates the maximum permittivity and conductivity of the mix.

Topp Equation

+

Figure: 3-13 Variation of dielectric constant and velocity with volumetric water content applicable in the GPR pla-teau frequency range (see Chapter 4).

These relationships are referred to as mixing relationships. There are a wide variety of theoretical and empirical mix-ing relationships given for materials. The complexity can be quite enormous depending on the nature of the problemto be addressed. Not only are the material interactions important, some of the interactions have finite response times,which lead to variation of electrical properties with frequency.

K a 3.03 9.3θv 146.0θv2

76.6θv3

– + +=

θv 5.3 102 –

2.92 102 – K a× 5.5 10

4 – × – +× K a

2=

4.3 106 –

× K a3

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



49

Archie's Law

E xample of Archie's law for a soil with 1000 ppm TDS groundwater and a surface contribution of 3 mS/m.

Figure: 3-14 Summary of Archie’s law which relates bulk conductivity to pore water conductivity and porosity.

The Topp equation (Topp et al, 1980) illustrated in Figure 3-13 is an empirically derived relationship which predictsthe bulk dielectric constant of soils as a function of volumetric water content. This relationship is most applicable inthe 50 MHz to 1000 MHz frequency range. A similar type of empirical relationship applies for conductivity versuswater content and its called Archie's Law (Telford et al 1976).

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



50

The form of Archie's Law shown is slightly different from the standard one which was originally proposed for use inthe petroleum industry at low frequencies of excitation. At GPR frequencies (10-1000 MHz), the pore water is themain contributor to bulk conductivity. An additional mechanism present is surface conduction. Surface conductionis associated with charges which are trapped on the surface of mineral grains in the composite material. The surfacearea per unit volume in a soil or composite material increases as the grain size of the material decreases. The amountof surface conduction depends on available surface area and the shape of the grains. Generally as grain size becomesfiner, as occurs in silts and clays, the electrical conductivity associated with surface conduction contributionsincreases. Mineralogy also plays a role as evidenced by mineralogical clays which have very large cation exchangecapacities (another indicator of large surface area) and exhibit strong surface conduction. As a result, fine grainedclay and clay-like materials (such as cement) tend to show higher conductivity at radio frequencies.

Another mixing formula, which is commonly used, is CRIM, which stands for Complex Refractive Index Model(Wharton et al, 1980). This is a totally heuristic mixing formula which weights the relative permittivity based on thevolume fraction and the square root of the complex permittivity of the material.

(3-15)

where K i and θi are the individual component permittivities and volume fractions.

While heuristic it can be quite helpful for a first order analysis of a problem when trying to get a handle on what one

might expect as the changes in permittivity.Another relationship is the Bruggeman Hanai Sen (BHS) model. This approach uses effective media theory to derivea composite material property from constituents (see Endres and Knight (1992)). Again, there are some basicassumptions about what comprises the host material and what is the additive material when dealing with multiphasemedia.

3.7 ELECTRICAL PROPERTIES OF WATER

Water is the single biggest factor which determines the bulk electrical properties of materials in most Earth settings.Water is present in all materials to some degree. The reason that water is important is that water is intrinsically a polar molecule. This means that the water molecule has a built-in net displacement of positive and negative electivecharges in its molecular structure as schematically depicted in Figure 3-15.

Figure: 3-15 Water plays an important role in determining velocities and attenuation in geologic materials.

∑=i

iimix K K θ

8/20/2019 GPR Annan 2003



51

Water is highly polarizable because the dipole moment of the water molecule will align itself within an applied elec-tric field. In other words, the dipole distribution in liquid water will redistribute itself to try to align with the appliedelectric field. This results in a relative permittivity for water in the range of about 80. The permittivity is temperaturedependent. Detailed discussions on the properties of water can be found in the informative book by Hasted (1972).

With water being a polar molecule, it will dissolve ionic materials. When ionic substances are placed in water theydissociate forming positively and negatively charged ions which conduct electricity by being mobile in the water.

Pure water is a poor electrical conductor. Conductivity is proportional to the numbers of dissolved ions present.Salinity and total dissolved solids (TDS) are measures of dissolved ion concentrations in water.

Water behaves as a polarizable material until frequencies exceed a few thousand MHz. Typically in the range of 10,000 MHz, the water molecule dipole moment can no longer track the applied electric field. The water moleculedipole moment rotation or alignment with an external exciting field is not synchronized which causes energy dissipa-tion in the material. This phenomenon is referred to as a natural relaxation process.

This water relaxation accounts for the utility of microwave ovens. Microwave ovens depend on water being presentin the material to be heated. The water relaxation dissipates the electromagnetic energy of the microwaves as heat inthe material.

Figure: 3-16 Variation of fresh water permittivity and conductivity with frequency illustration the water moleculedipole moment relaxation response in the 10,000 MHz frequency range.

The relationship between permittivity and conductivity in pure water with frequency is depicted in Figure 3-16. Theimplications for GPR are that as the GPR frequency exceeds the 500 to 1000 MHz range, the presence of water becomes more and more important as dissipation of energy increases substantially as frequency increases. As aresult, only shallow depth sounding applications are practical with GPR at frequencies much above 1000 MHz unlessmaterials are dry (water free). More discussion on frequency dependent properties will be addressed later.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



52

3.8 R EAL GPR EXAMPLES

3.8.1 DNAPL SPILL

To illustrate the concept of working with composite materials, two applications of GPR to ground water problems are presented. The first example is the issue of locating DNAPL (acronym for dense non-aqueous phase liquid). ADNAPL represents a hazardous contaminant in the subsurface. A DNAPL when spilled on the ground sinks down

into the subsurface. Because it is denser than water, it will tend to displace water from the pore space and sink downthrough water saturated soils.

DNAPL's are not soluble in water to any large degree. They can be toxic (in some situations) at the part per billionlevel but soluble at the part per million level. A small pond or pool of DNAPL in the subsurface can contaminate anaquifer for drinking purposes for a substantial period of time.

The present example describes a controlled experiment which was carried out to determine whether GPR (and other electric and electromagnetic techniques) could be used to detect the presence of DNAPL's in the subsurface and totrack their migration. The experiment is extensively described by Greenhouse et al. (1992) and Brewster et al.,(1995).

In this experiment a sandy aquifer at surface was confined using sheet pile walling and a clay aquatard at about a 3 mdepth to form a natural cell for the experiment. The composite natural material consisted of beach sand, with someother organic soil present, water, air and DNAPL. For this particular problem the water table was held at surface sothat the whole section was water saturated to minimize the air component in the composite mixture. The medium canconsidered as a three component mix of soil grains, water and DNAPL.

The soil grains are essentially non-conducting with a permittivity of about 3 to 4. The electrical properties are con-trolled by the presence of water which, given the soil porosity of about 30 to 35%, creates a bulk permittivity of 25for the water saturated soil. Similarly, the bulk conductivity is controlled by the ground water conductivity. The bulk saturated soil conductivity is 15 mS/m.

The DNAPL is an insulator with a permittivity in the range of 1.5 and does not conduct on electricity. When it entersthe soil, it displaces water from the pore space. Any zone which is substantially penetrated by DNAPL, exhibits areduction in both permittivity and conductivity. Experimental evaluation of the relationship between electrical prop-erties and DNAPL saturation in the water-saturated material are graphed in Figure 3-17. As the DNAPL displaceswater, the permittivity and the conductivity both decrease. These changes in electrical properties define the effective-ness of the experiment.

Relative Dielectric Permittivity Conductivity vs PCE Saturation Soil/PCE/Water Mixture - BHS Model Lab Observation

40% porosity4.5/2.3/83

(a) (b)

Figure: 3-17 Illustration of how permittivity and conductivity change when DNAPL replaces water in the pore space of a water saturated sandy soil.

8/20/2019 GPR Annan 2003



53

GPR measurements were made at regular time intervals on a grid starting prior to DNAPL injection. The DNAPL penetrated into the subsurface and spread outwards and downwards with time. Changes in the GPR reflectionresponse with time indicate areas of DNAPL concentration build up. As the DNAPL pooled in zones it generatedstrong reflections of the GPR signal. Figure 3-18a shows the GPR response prior to DNAPL injection. The responseat three times after the start of injection are shown in Figure 3-18a, b, c, and d.

(a) Cell line E5 prior to spill (b) Cell line E5 at 54 hours

(c) Cell line E5 at 177 hours (d) Cell line E5 at 235 hours

Figure: 3-18 Time lapse GPR cross-sections of an experimental DNAPL spill showing that high concentrations of DNAPL modify electrical properties giving rise to GPR reflections.

The first feature of importance is the strong reflection associated with pooling of the DNAPL at 1 m depth. Second,deeper natural events are affected because the travel time through the DNAPL saturated zone is reduced because of the higher velocity. Third, the reflection amplitude from the aquatard at the bottom is reduced because energy isreflected from the shallower DNAPL pool. These responses are typical manifestations of what one would see at areal site and it illustrates the importance of understanding the interaction of the various constituents in the overallcomposite mix when evaluating the GPR response.

3.8.2 CONDUCTIVE CONTAMINANT

The second example is that of a leachate contaminate from a landfill percolating through a ground water system. Inthis case, the landfill has been active for a number of years and the area of testing was in a sand and gravel pit. Thesand and gravel have been removed down to the depth of the water table and the pit has been abandoned. The landfillis located up ground water flow direction and generates a leachate. The leachate liquid is chloride ion rich which boosts the pore water conductivity when leachate is present.

8/20/2019 GPR Annan 2003



54

The survey area soil consisted of clean sand of glacial/fluvial deposition. The sand is transparent to GPR signals andtypically 20 meters of these materials can be penetrated throughout the survey area with one exception. In the areadirectly down ground water flow direction from the landfill, the radar signals failed to penetrate to the same depth.This effect is caused by changes in the pore water conductivity due to the presence of leachate contaminant from thelandfill. An example GPR section down is shown in Figure 3-19. These data clearly show the effect of the leachate-induced change in conductivity of the pore water.

Figure: 3-19 Leachate from a landfill modifies pore water conductivity. The GPR section over the same line at different times illustrate how electrical conductivity affects GPR signals.

From a geologic and a visual observation perspective, the area is uniform; in other words there is water saturated sandthroughout the whole survey area. The only thing that changes is the chemistry of the water. The presence of thechloride ion rich leachate boosts the conductivity of the pore water causing the radar signals to be more stronglyabsorbed and hence the signal penetration depth decreases. This example can be understood using Archie's Law withvariable pore water conductivity as the factor controlling the composite mix conductivity.

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 4-EM Wave Properties

55

4 EM WAVE PROPERTIES

4.1 WAVEFRONTS AND R AYS

Electromagnetic fields when propagating as waves can be characterized by wavefronts or by ray paths. Bothconcepts are depicted in Figure 4-1. The wavefront is the spatial surface on which the signals are all in phase. For a

transient signal emitted by a localized source, the equal travel time spatial surface defines the wavefront. The

definitions are actually quite a bit more exacting if one studies the subject of geometrical optics (Born and Wolf,

1980) but this simple description suffices for our purposes.

Figure: 4-1 wavefronts are surfaces of equal travel time or phase for waves travelling out from a source. Rays are perpendicular to the wavefronts and sketch out the from trajectory. Both concepts are useful in visualizing GPR sig-nals.

Ray paths are most appropriate when the wavelength or temporal signal duration is very short. Ray paths are

perpendicular to the wavefront. The electromagnetic fields when treated as rays conceptually travels along the path

defined by the ray. The electric and magnetic fields are perpendicular to the ray path.

Wavefronts and the ray paths are very helpful in visualizing EM fields as they move through a medium. For a more

in depth discussion of these topics, the text by Born and Wolf is very helpful; the formal mathematical foundations of

these concepts are well presented.

4.2 WAVE PROPERTIES

Wave properties were developed earlier in chapter two. The key EM wave field properties are phase velocity, v,

attenuation, α, and electromagnetic impedance Z. The behaviour of these wave properties for a simple medium with

fixed permittivity, conductivity and permeability are illustrated in Figure 4-2.

Wave fronts

Rays

source

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003


4-EM Wave Properties GPR Principles, Procedures & Applications

56

(a)

(b)

Figure: 4-2 Variation of velocity and attenuation in a simple medium.

The wave properties all show similar behaviour. At low frequencies, all properties depend on √ω. At high

frequencies, the properties become frequency independent if are constant. The high frequency behaviour is

the character of most importance to GPR.

At low frequencies, the electromagnetic fields diffuse into the material. The energy distributes itself in the same

manner as heat. An impulsive signal gets smeared out (dispersed) because its frequency components are attenuated at

different rates and travel at differing phase velocities. The mathematical form for phase velocity, v, attenuation, α,

and electromagnetic impedance Z are

(4-1)

“DIFFUSION”

“DISPERSIVE”

“PROPAGATION”

“NON-DISPERSIVE”

2

f T

log frequency

l o g

a t t e n u a t i o n

K

1

2

Zo

“DIFFUSION”

“DISPERSIVE”

“PROPAGATION”

“NON-DISPERSIVE”

2 f T

c

K

log frequency

o g

v e

o c

t y

ε µ σ, ,

σ µ

ω

⋅⋅

=2

v

8/20/2019 GPR Annan 2003



57

(4-2)

(4-3)

At high frequencies, the electromagnetic fields propagate as waves through the medium. All frequency components

travel at the same velocity and suffer the same attenuation. An impulsive signal will travel with its shape intact. This

is referred to propagation without dispersion (see Annan, 1996). The velocity, attenuation and impedance are

expressed as

(4-4)

(4-5)

(4-6)

with the right most expression being valid where magnetic properties are assumed to be unimportant. Z0 is the

impedance of free space.

(4-7)

The transition from diffusion to propagation behaviour occurs when the electrical currents change from conduction

dominant to displacement current dominant behaviour. For a simple material, the transition frequency is defined as

(4-8)

In real materials, the electric and magnetic properties can be frequency dependent as illustrated for water in chapter 3.

In this case there will always be a tendency for permittivity to decrease with frequency and conductivity to increase

with frequency.

2

σ µ ω α

⋅⋅=

( ) σ

µ ω

⋅

⋅

+= 2i1Z

K

c1v =

⋅=

µ ε

K ⋅⋅=⋅=2

Z2

0

σ σ

ε

µ α

K

ZZ 0==

ε

µ

ohms377Z0

00 ⋅==

ε

µ

πε

σ

⋅=

2f

8/20/2019 GPR Annan 2003



58

4.3 GPR PLATEAU

Real materials generally behave more as depicted in Figure 4-3. The wave properties will tend to show a plateau,

called the GPR plateau, at some frequency range. In some materials this plateau may not exist. Such media are not

amenable to traditional GPR measurement methods. The fields will be diffusive in character requiring interpretation

of observations using traditional transient electromagnetic induction approaches.

Figure: 4-3 In most geological materials, electrical properties at high frequency will be affected by the presence of

free water in the pores and fractures. The water relaxation of 10 GHz causes increased effective conductivity andreduced permittivity. Velocity, attenuation and impedance for materials exhibit a plateau where GPR is a most appli-cable method.

The “plateau” may still exhibit some gradual increase in velocity and attenuation with frequency. The increase in

attenuation is usually the most important as many GPR applications are close to the attenuation limit and any increase

can mean the difference between success and failure of GPR. There are two primary factors which induce this

increase. One is the presence of water. As discussed in chapter 2, water starts to absorb energy more and more

strongly as frequency increases toward the water relaxation frequency of 10,000 MHz. Even at 500 MHz water

losses can start to be seen in otherwise low loss materials.

8/20/2019 GPR Annan 2003



59

The second factor is scattering loss which are discussed later in this chapter. Scattering losses are extremely

frequency dependent and can become more important that electrical losses in many instances.

4.4 MATERIAL INTERFACES

GPR is most frequently used to detect and map features at a distance. To be detectable, the object in question mustemit signals. The mechanism for creating such signals is a difference in the electrical properties of the object from

those of the surrounding material.

The reflection, refraction and transmission of electromagnetic waves is discussed in the basic EM texts cited

previously. The key issues are addressed in the following sections with the concepts most important to GPR and the

interpretation of the results.

4.5 SNELL’S LAW

Snell’s law expresses how wavefronts change direction as the fields move through materials where velocity is not

constant. The concept goes back to the earliest geometrical optics and the bending of light rays. The same lawapplies with seismic waves.

Figure: 4-4 Snell’s law predicts that waves change direction of propagation when the waves encounter changes inmaterial properties

The concept is illustrated in Figure 4-4. An EM signal is incident on the boundary between two materials of differing

properties. In order for the fields to match at the interface, the horizontal (x direction) movement of the phase must

match. From chapter 2, the basic building block is the plane wave and it suffices to illustrate Snell’s law. A plane

wave incident on the boundary from medium 1 has its propagation vector at an angle θ1 to the vertical (z axis). When

the fields traverse through the interface, the direction of the propagation vector must change. In medium 2, the propagation vector makes an angle θ2 with the vertical (z axis). Mathematically, Snell’s law requires the horizontal

component of the propagation vector in each material to be equal.

(4-9)

When the materials are low loss k=ω/v, Snell’s law takes the more simple form

Medium 1

Medium 2

k 1

k2

1θ

2θ

2211 sinsink θ θ ⋅=⋅ k

8/20/2019 GPR Annan 2003



60

(4-10)

If multiple interfaces are present, then this rule must be applied at each boundary and the wavefront or rays must bend

or change direction at each boundary. If the velocity or materials vary continuously, this relationship still applies.

In GPR, medium 1 can readily be envisioned as the air environment and medium 2 envisioned as the ground or the

material to be probed. Since air velocity is always substantially faster than other materials, all signals from the air

will bend and travel more vertically downward.

4.6 CRITICAL ANGLE

The critical angle plays a very important role in GPR. It appears in many discussions and interpretations of GPR

responses. From the discussion in the preceding section, Snell’s law requires the wave fields to bend or change

direction in a prescribed manner as velocity changes.

If v2<v1 at a simple interface, there a range of angles θ2 which can not be illuminated from medium as depicted inFigure 4-5.

Figure: 4-5 When signals travel from a high velocity (material 1) to a lower velocity (material 2), impinging signals

for angles from 0 to 90o only emerge in material 2 between cingles 0 and sin . is normally referred to as the

critical cingle.

Incident signals can impinge on the interface between the media at any angle θ1 from 0 to 90°. The maximum of angle

θ2 is limited to

. (4-11)

This value of θ2 is called the critical angle θc. When a signal travels horizontally in medium 1, the signal is refracted

downward at the critical angle. In simple ray optics analysis, propagation at angles θ2 <θc cannot be energized from

2

2

1

1

v

sin

v

sin θ θ =

varies from

0 to 90o

θ1

θ2

varies from

0 to θc

Material 1

Material 2

Range of angles

not reachable in

the classic sense

θc θc

1

2

2v

vsin =θ

8/20/2019 GPR Annan 2003



61

medium 1. As we shall see later, this concept is not completely valid when we deal localized sources near the

interface.

When signals travel upward from medium 2 at angles greater than θc then these fields require that the angle of

propagation in medium take on an counter intuitive character. The details will become more apparent as we deal with

waves in layered media later in this chapter. When θ2 =θc, then θ1 = 90°. The fields in medium 1 travel horizontally.

When θ2 >θc then sin θ1 must be greater that 1 which it not possible for normal real physical angles. Mathematically

sin θ1 can be greater than 1 if θ1 is a complex number. Physically this means that the horizontal phase velocity must

be less than the phase velocity of the material and is only possible if the fields decay exponentially away from the

interface in medium 1. Such a wave is called an evanescent wave and only exists at the boundary between materials

or in the vicinity of finite field sources.

4.7 FRESNEL R EFLECTION COEFFICIENTS

The Fresnel reflection (and transmission) coefficients quantify how the amplitudes of the electromagnetic fields vary

across an interface between two materials. Since this topic is treated exhaustively in every EM and optics texts

published for the past 100 years, only the essential physics will be reviewed here and the mathematical form

presented. See the texts by Jackson (1967) or Born and Wolf (1980) on this topic for details on the mathematical

development.

When a plane EM wave impinges on a boundary, it is partially transmitted and partially reflected as depicted in

Figure 4-6.

Figure: 4-6 Waves incidence on a planar interface are partially transmitted and reflected. Here the incident signalwith amplitude I give rise to a reflected signal RI and a transmitted TI.

The amplitude of the incident field is denoted as I and the reflected signals are denoted as R I and T I where R and T

are the transmission and reflection coefficients. At this juncture, we need to get specific about the nature of an EM

wave. As we saw in chapter 2, for a given direction of propagation, there can be two independent waves.

When planar boundaries are encountered, it has become traditional to decompose an incident wave into two

components whose vector components have a compatible orientation with respect to the boundary. These two wave

fields are referred to the TE (transverse electric field) and TM (transverse magnetic field) waves and are depicted in

Figure 4-7. The TE wave always has its electric field parallel to the same plane of the interface while the TM wave

X

1

2

1

IRI

TI

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



62

has its magnetic field in the plane of the interface. This decomposition is strictly a function of the interface geometry

and has nothing to do the EM fields specifically. By decomposing the field into TE and TM components, the specific

mathematical form of R and T can be derived.

Figure: 4-7 EM waves are transverse vector waves field. For any given propagation direction, two independent fields exist. For planar interfaces, it is tradition to discuss the two waves; one with the electric field in the interface plane called traverse E (TE) and one with the magnetic filed vector in the interface plane called traverse magnetic

(TM).

The reflection and transmission coefficients for TE and TM waves take different mathematical forms because thefield behaviour is different. Formally

(4-12)

and

(4-13)

where ITE represents the electric field strength for the TE wave and ITM represents the magnetic field strength for the

TM wave.

The mathematical forms of R and I are derived by noting two fundamental facts. First, Snell’s law must be satisfied.

Second, physical behaviour requires electric and magnetic fields in the plane of the interface to be the same on either

side of the boundary plus the electric current and magnetic flux density crossing the boundary must be the same on

either side of the boundary.

When these conditions are met one finds that

Incident

FieldReflected

Field

E

H

E

EH

H

Transmitted

FieldInto page

Out of page

Transverse ElectricField

(TE)

Transverse MagneticField

(TM)

Incident

FieldReflected

Field

E

H

E

EH

H

Transmitted

Field

TETETETETE ITIR I ⋅=⋅+

TMTMTMTMTM ITIR I ⋅=⋅+

8/20/2019 GPR Annan 2003



63

(4-14)

(4-15)

and

(4-16)

(4-17)

are the forms for the reflection and transmission coefficients.

It is important to remember that the above expressions apply to the magnetic field in the TM case and the electric

field in the TE case. In some instances, the behaviour of the magnetic field in the TE case and the electric field in the

TM case are of interest. The reflection and transmission coefficients for the field components are expressed as

follows where the superscripts indicate the field component.

(4-18)

(4-19)

(4-20)

(4-21)

When the EM wave is vertically incident on the interface (θ1 = 0°), there is no distinction between a TE and TM wave

and the TE and TM reflection coefficients become identical (for the field components). For none vertical incidence,

the coefficients are different. Two examples are shown graphically in Figure 4-8 for the case when electrical losses

are small enough to ignored. In the first, the velocity decreases at the interface while in the second the velocity

increases. Behaviour impedance contrasts for 2:1 (dry soil) to 9:1 (water) are depicted.

2211

2211TE

cosYcosY

cosYcosYR

θ θ

θ θ

⋅+⋅⋅−⋅

=

2211

2211TM

cosZcosZ

cosZcosZR

θ θ

θ θ

⋅+⋅

⋅−⋅=

2211

11TETE

cosYcosY

cosY2R 1T

θ θ

θ

⋅+⋅⋅⋅

=+=

2211

11TMTM

cosZcosZ

cosZ2R 1T

θ θ

θ

⋅+⋅⋅⋅

=+=

E

TE

H

TE R R −=

H

TM

E

TM R R −=

E

TE

1

2H

TE TY

YT ⋅=

H

TM

1

2E

TM TZ

ZT ⋅=

8/20/2019 GPR Annan 2003



64

TE TM

Figure: 4-8 TE and TM reflection coefficients as a function of incidence angle.

Theses results show the 4 important points that should be remembered when evaluating and interpreting GPR data.

First, the reflection magnitude becomes larger at large angles.

Second, the TM reflection coefficient can show a null or a reduction to a minimum as the angle of incidence

increases. The angle of the minimum is known as the Brewster angle. At the Brewster angle, maximum transmission

through interface occurs. For TE waves, the admittance must decrease at the interface for Brewster angle to exist; for

TM waves, the impedance must decrease crossing the interface.

Third, when the waves are travelling from a low velocity to higher velocity medium, the magnitude of the reflection

coefficients becomes unity for angles greater that the critical angles. The waves are totally reflected but as mentioned

in the previous section, fields do exist in the other material but behave as evanescent signals which decay

exponentially with distance from the interface.

Four, the sign of the reflection coefficients can be either positive or negative. A positive reflection coefficient means

the reflected field (E for TE and H for TM) are in the same direction as the incident field vector while a negative

coefficient means the reflected field is in the opposite direction to the incident field direction such as depicted inFigure 4-9.

Figure: 4-9 Illustration of polarity of signals for normal incidence. Note R can be either positive or negative whichcan prove a very powerful interpretation diagnostic.

8/20/2019 GPR Annan 2003



65

In order to provide a frame of reference, the normal incidence reflection coefficients for some common interfaces that

might be encountered in GPR are tabulated in Table 4-1.

Table: 4-1 Example of normal incidence reflection coefficient for some common GPR

interfaces assuming conductivity is negligible.

4.8 THIN LAYER R EFLECTION

A common GPR target is a layer of different material sandwiched in a uniform background. There are many

examples – a crack in concrete, a fracture in rock, a thin clay layer in a sand deposit, a DNAPL spreading an

impermeable soil horizon. The question that often arises is “Can a layer of a given thickness be detected?”. Since the

mathematical solution is relatively simple, it is provided here as a guide for many later discussions.

Figure: 4-10 When a signal is incident on a layer embedded in a medium, signals are partially transmitted andreflected. Signals entering the layer bounce back and forth between the interfaces re-emitting reflected and transmit-

ted signal replicas which occur at fixed time delays (or with fixed phase shifts).

Boundary K 1 K 2 Z1 Z2 R

Air-dry soil 1 4 377 188 - 0.05

Air-wet soil 1 25 377 75 - 0.67

Dry soil - wet soil 4 25 188 75 - 0.43

Dry soil - rock 4 6 188 154 - 0.01

Wet soil - rock 25 6 75 154 + 0.34

Ice - water 3.2 81 210 42 - 0.67

Moist soil - water 9 81 126 42 - 0.5

Moist soil - air 9 1 126 377 +0.5

Soil - Metal 9 126 0 -1∞

T1 T2 T3

R1 R2 R3 R4

Material 1

Material 2

Material 1

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



66

Figure 4-10 depicts the nature of the problem and how the response can be heuristically derived by following ray

paths. The response is not a single event but a host of events representing the reverberation of the signal back and

forth in the layer. Mathematically, the reflection coefficient is expressed

(4-22)

where R pq and T pq are the previously defined fresnel reflection and transmission coefficients at the boundary between

medium p and medium q. The factor β represents the time delay or the phase shift the signal encounters on travelling

down and back through the layer. For a normal incident monochromatic (i.e. sinusoidal excitation) wave

(4-23)

where k 2 is the propagation in the layer and d is the layer thickness.

A similar formulation can be developed for the transmitted signal but this will not be explored here.

This expression for R is just a geometric series which can be summed with the result that

(4-24)

To understand the behaviour of this reflection coefficient, the result for zero loss as of function of sinusoidal

excitation frequency is depicted in Figure 4-11a). The horizontal axis is since

(4-25)

where f is excitation frequency, v2 is the velocity in the layer and x2 is the wavelength in the layer.

a) b)

Figure: 4-11 The embedded thin layer reflection coefficient magnitude for sinusoidal excitation. The reflectionamplitude varies as the ratio of layer thickness to wavelength. For a lossless layer the magnitude oscillates between

0 and a maximum value. If the layer has some loss, the oscillations damp out as the layer becomes thick.

R R 12 T12T21R 21β T12T21R 213 β2

T12T21 R 215 β3 …+ + + + +=

β ei2k 2d

=

R R 12

T12T21R 21β

1 R 212

β – -----------------------------+=

d λ2 ⁄

k 2d

2πd

λ2----------

2πfd

v2------------= =

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



67

The reflected response goes to zero at zero frequency and oscillates with maximum at

(4-26)

where n is an integer. The maximum value is

(4-27)

and minimum is

(4-28)

If the layer has some finite high-frequency loss, R has less extreme variations that die out as increases as shown

in Figure 4-11 b). The maxima and minima just indicate when the reflections from the top and bottom of the layer add

together constructively or destructively.

Of interest in many situations is the behaviour when is small. In this case

(4-29)

and

(4-30)

Regrouped one has

(4-31)

The expression has two important consequences. First, R depends linearly on the layer thickness which has many

uses in the interpretation of GPR responses. Secondly, for traditional finite bandwidth GPR systems, all the

excitation frequencies in the pulse can be such that k 2d << 1. In other words, the excitation spectrum of the

incident signal is small (zero) when k 2d gets large. In this case, the transient response of the thin layer has the form

d λ2 ⁄ 0→( )

d

λ2

-----1

4---

n

2---+=

R R 122

1 R 122

+------------------

=

R 0=

d λ2 ⁄

d λ2 ⁄

β 1 i2k 2d+≈

R R 12

1 R – 122

-----------------i2k 2d≈

R R 12

1 R 122

– -----------------

2d

v2

------ –

iω=

S ω( )

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



68

(4-32)

where is the band limited excitation and C is the bracketed component of Equation (4-31). When transformed

to a temporal response, the reflected signal becomes

(4-33)

In other words, the temporal shape of the reflected signal is the time derivative of the incident pulse multiplied by a

constant that depends linearly on layer thickness. The result is depicted graphically in Figure 4-12.

Figure: 4-12 When a layer is very thin compared to all wavelengths in a pulse, the reflected signal looks like thetime derivative of the incident signal.

The thin (and finite thickness) layer response discussed here contain many of the essential characteristics of all GPR

targets. The simple model is also useful in addressing the issue of resolution versus detection of a feature. Resolution

requires layer thickness to be on the dimension of a wavelength or the travel time through the layer to be

comparable or greater than the pulse duration. Detection only requires the thin layer reflection to be detectable above

the noise level.

4.9 PLANE WAVE MODEL FOR FINITE SOURCES

One of the critical aspects of GPR is the behaviour of the antennas when placed on the ground. This topic can

become quite mathematical to quantify the behaviour in even the simplest cases. Again, we will follow the heuristic

R ω( ) iωCS ω( )=

S ω( )

r t( ) Cd

dt-----s t( )=

Reflected

Signal

Incident

Signal

λ2

( )

8/20/2019 GPR Annan 2003



69

approach giving the essential results without resorting to full mathematical formalism.

Figure: 4-13 A spherical (in general curved) wavefront can be synthesized as a superposition of planar wavefronts.

The behaviour of a small (by which we mean d/v<<1 in all materials where d is the source dimension) current

element can be treated as a source of spherical wavefronts as we discussed in chapter 2. The spherical wavefront can

be envisioned as being built up of planar wavefronts which are tangential to the spherical wavefront as depicted in

Figure 4-13. The spherical wave S is thus expressed as

(4-34)

where f n are the relative wave amplitudes. In the simple two dimensional case, the plane waves travel in the

directions between 0 and 2 and Pn have the form

(4-35)

where is a unit vector in the direction .

Writing out as its geometric components.

P0

P1

P2

P3

P4

PN

... . . .

S f 0P0 f +1P1 f +

2P2…=

N

Σn 0=

f nPn

θπ

Pn eikr d θ

n( )⋅

eik z θ

n

x θsin+n

cos( )= =

d θn i2π

N 1+-------------

=

kd

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



70

(4-36)

where

(4-37)

While the derivation here is simplistic, in fact the outgoing fields from a localized source can be represented with full

accuracy with the limit of letting and letting k x range from to . The mathematical form becomes

(4-38)

which is just a spatial fourier transform for those familiar with fourier analysis. When the full 3D problem is

addressed we have to introduce the y dimension and the resulting form is

(4-39)

where

(4-40)

This basic type of mathematical formalism has been known and used for over a century. One can refer to the work of

Sommerfield (1949), Banos (1962), Wait (1962) and Brekhovskikh (1962).

From this formalism, although very simplified, it is possible to understand the first order behaviour of a GPR system

which is addressed in the next section.

A slight of hand here was to allow k x to range from to instead of just –k to k. Physically this means that the

angle of incidence range is beyond the physically observable range of 0 to /2 (90o). If one carries out the full

formal solution, one must mathematically include waves for which k x > k. The meaning of such a wave is seen by

noting that k z becomes imaginary (complex or square root of a negative number). This rather odd result is called an

evanescent wave. The 2D plane evanescent wave has the form

Pn ei k z

nz k x

nx+( )

=

k zn

k 2

k xn2

– ( )=

k xn

k θnsin=

N ∞→ ∞ – ∞

s x z,( ) F k x( )eik

xx

eik

zz

k xd

∞

∞

∫=

s x y z, ,( ) F k x k y,( )ei k xx k yy k zz+ +( )

k xd k yd

∞

∞

∫∞

∞

∫=

k z k 2

k x2

– k y2

– ( )1 2 ⁄

=

∞ – ∞π

8/20/2019 GPR Annan 2003



71

(4-41)

and, if k x > k, the and the plane wave is expressed

(4-42)

A reference horizon is always involved (i.e. z = source position). The sign of z dependence is picked to yield a

physically correct field (that is one where the field decays away from the reference plane). The result is sketched in

Figure 4-14.

Figure: 4-14 An evanescent wave is one tied to a reference surface. The wave propagates parallel to the surfaceand decays exponentially with distance away from the surface.

4.10 GPR SOURCE NEAR AN INTERFACE

To continue with our exploration of GPR sources over the ground, the afore discussed wave model describing fields

spreading out from the source leads to the visualization in Figure 4-15. When our localized source is above the

ground, the signals spread out as depicted in Figure 4-15. The spherical wavefront impinges on the ground. Ignoring

finite source dimension and wavelengths, the field at any point along the ground interface can be visualized locally as

a planar wave impinging on the boundary at a specific incidence angle defined by geometry (source height and lateral

distance). Locally the signal is reflected and refracted according to Snell’s law and the fresnel reflection coefficient.

ei k xx k zz+( )

k z i k x2

k – ( )1 2 ⁄

±=

ek x

2k

2 – ( )

1 2 ⁄ z± ik xx+

Propagation

direction

Attenuation

direction

Support

boundary or

interface

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



72

(a)

(b)

Figure: 4-15 wavefronts spreading out from a localized source a) located above the ground and b) located on theair-ground interface.

If one examines the wavefront in the ground, it is no longer spherical as bending occurs with differing degrees

depending on the varying incidence angle. To understand what is happening in the ground and near the interface, the

limiting case of the source right at the interface is informative (see Figure: 4-15b). The incident and reflected waves

in the air coalesce into an upgoing spherical wave. In the ground, the transmitted signal divides into two parts, a

spherical wave and a planar wavefront travelling at the critical angle which links the direct spherical air wave and the

spherical ground wave. Near the interface, the spherical ground wave extends into the air as an evanescent field.

The derivation of the mathematical form is based on the plane wave field expansion discussed in the previous section.

For those interested, references such as Sommerfield (1949), Wait (1962), Brekhovskik (1960) and Annan (1973)

have discussions. The various wave fields are clearly visible at large distances from the source and/or very short

wavelengths. For short distances from the source or long wavelengths, the separation of the events is blurred but the

essential concepts are valid.

For a localized target in the ground, there can be several possible paths that energy can travel from a transmitter to the

receiver. The concepts are depicted in Figure 4-16 and the ray paths are shown in Figure 4-17.

Figure: 4-16 Concept drawing of a localized target beneath the ground with a transmitter and receiver located atthe air ground interface.

Air

Ground

Critical Angle

Air

Ground

T R Air

Ground

Local Target

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



73

a) Direct air signal b) Direct ground signal

c) Direct reflected (scattered) signal d) Transmitted signal criticallyrefracted and target scatters

towards receiver.

e) Transmitted signal scattered and f) Critically refracted transmittedcritically refracted to receiver. signal is scattered and critically

refracted to the receiver.

Figure: 4-17 For the simple model in Figure 4-15 , there are many paths that signals can travel between a transmit-ter and a receiver.

The relative importance of each path depends on the target depth, the separation between the transmitter and receiver

and elevation of the transmitter and receiver. One should also note that one can not distinguish between d) and e).

In most GPR cases, the transmitter receiver separation is small and the predominant paths are a), b) and c). Paths d)

and e) can still be important if both the transmitter and receiver are a substantial distance from the target even if the

transmitter and receiver are close together.

The wavefronts are most commonly represented as ray paths as these are simpler to draw. For the source on theground, the ray paths to a receiver at separation x and associated signal travel times are depicted in Figure 4-18.

T R

Local Target

T R

T R R

T R T

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



74

Figure: 4-18 The wavefronts about a source on the ground ( Figure 4-15 ) are often represented as ray paths asdepicted. d is the layer 1 thickness, v is the layer 1 velocity and c is the speed of light in the air. The arrival time of

these rays versus source-receiver separation provide a means of determining ground velocity structure.

As indicated earlier, the intent here is to provide a basic overview. For the curious, dig into the references. Many of

the ideas will be revisited in later discussions where some aspects are explored in more detail.

4.11 R ESOLUTION AND ZONE OF INFLUENCE

Since GPR detects objects at a distance, the question that always arises is how accurately can the object be locatedand what detailed information can be extracted about the geometry of the object. Resolution indicates how preciselythe position can be determined. Closely related to the question of resolution is the question of what geometricalattributes of the target can be extracted. Geometrical attributes include such factors as the size, shape, thickness, etc.

Resolution divides into two topics; there is longitudinal (range or depth) resolution and there is the lateral (or angular)resolution. The basic concepts are depicted in Figure 4-19.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



75

Figure: 4-19 Resolution for GPR divides into two parts; namely range resolution and lateral (or angular) resolu-tion.

Understanding resolution gets into the fundamental issues of GPR detection concepts. These concepts are also

common to seismic measurements and in fact they are applicable to any technique where wave phenomenon are used

to detect objects at a distance.

In the current discussion we will work in the time domain. We will be considering systems which generate a pulse

and detect the echoes from distant targets. Echoes may arrive simultaneously, overlap or be separated in time as

depicted in Figure 4-20.

Figure: 4-20 Temporal pulses with 1/2 width of W. Pulses are clearly separable when T>>W (a). Two pulses are

said to be distinguishable until (b). When T<<W then two events are not distinguishable (c).

T R

r

l

rangeresolution

lateral or angularresolution

Pulses

Coincident

T

Pulses

overlap

W

Pulses

Clearly

Separate

T

(a)

(b)

(c)

clearly

separate

Pulses

overlap

Pulses

coincident

Pulses

T W≈

8/20/2019 GPR Annan 2003



76

When two responses are present, how closely spaced in time can they be and still be discerned as distinct separate

events. If two pulses are coincident in time, the amplitude will be enhanced. The result will be one event with a larger

amplitude. As time separation between the events increase they can be recognized distinctly as two events when they

no longer overlap. This subject in raised again in chapter 5 when we talk about system bandwidth and resolution.

Generally speaking a pulse is characterized by its width at half amplitude, W. While there are many definitions of

resolution, the widely accepted definition of resolvable pulses requires that the two pulses be separated by half their “half width” in order to be distinguishable as two events. If they are separated in time by less than this amount then

they will most likely be interpreted as a single event.

These temporal pulse concepts must be translated into the spatial domain to define spatial resolution. The approach is

best understood by examining the response of two point targets as illustrated in Figure 4-21.

(a) (b)

Figure: 4-21 Range and lateral resolution can be determined by considering the response of two localized targetseither in-line (a) or side-by-side (b).

For two targets denoted 1 and 2, as shown in Figure 4-21, the longitudinal difference in travel time (which is

observed on a GPR record) is directly related to the difference in radial or longitudinal distance between the targets.

The travel time for the first target will be

(4-43)

and the travel time for the second target will be

(4-44)

where v is the propagation velocity of the signal.

∆ r

r

T R T R

r

∆ L

1

2

1 2

ll l∆

t12r

v-----=

t22r 2∆r +

v---------------------=

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



77

The differential time is expressed as

(4-45)

and this time difference must be greater than half the pulse half width in order that the responses be detected as twoevents. The spatial separation of targets in the radial direction from the system must be greater than or equal to

(4-46)

One can see from this analysis that the pulse width and the velocity in the material combine to dictate the radial

resolution. The radial resolution is essentially independent of distance from the source in an ideal world.

The lateral resolution is analyzed in a similar fashion. Following along from the geometry in Figure 4-21, the travel

time for target 1 is

(4-47)

and the travel time for target 2

(4-48)

The time difference between the two events is expressed as

(4-49)

In most situations the target is a substantial distance away from the measurement system permitting the use of an

approximation. When this approximation is employed, the time difference is the square root in Equation (4-49).

(4-50)

The lateral resolution (minimum separation of two side-by-side targets to be distinguishable) must be

∆t t2 t12∆r

v---------= – =

∆r Wv

4---------≥

t12r

v-----=

t2 2 r

2

∆l

2

+( )

1 2 ⁄

v----------------------------------=

∆t2 r

2 ∆l2

+( )1 2 ⁄

r – [ ]v

-----------------------------------------------=

∆t ∆l

2

vr --------≈

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



78

(4-51)

From this result one can see that the lateral resolution depends on the velocity and the pulse width as well as thedistance from the system. The larger the distance to the targets, the lower the lateral resolution.

The lateral resolution is closely related to the Fresnel zone concept which expresses the resolution concepts for

interference of monochromatic (sinusoidal) signals.

With GPR, the pulse width, W, in time is directly related to the bandwidth, B, which is normally related to the center

frequency, f c. Using the relationship

(4-52)

and

(4-53)

where is the wavelength of the GPR center frequency, one finds that the lateral resolution can be expressed as

(4-54)

This result is identical to the expression for the Fresnel zone for monochromatic signals.

Lateral resolution actually defines an area or zone of resolution since all targets encompassed by a radius of perpendicular to r can not be resolved.

∆lvrW

2-----------≥

W1

B

----1

f c----==

λc f c v ⁄ =

λc

∆ldλc

2--------=

21∆2

l∆

8/20/2019 GPR Annan 2003



79

Figure: 4-22 The zone of influence is defined as the area which can curtain a second target that can not be uniquelyresolved.

The zone of influence extends into a “tube” of influence when the concepts are extended to transillumination GPR

techniques. These concepts are not unique to GPR and an excellent discussion of these same concepts from a seismic

perspective is presented by Knoll (1991).

4.12 SCATTERING FROM A LOCALIZED OBJECT

Time and again with GPR, one has to address the issue of energy returned from a localized feature in the subsurface.Trying to quantify this behavior is complex and full analysis is beyond the scope of this introductory set of notes. In

general, one can visualize the problem as depicted in Figure 4-23. An electromagnetic field is incident on a localizedobject which gives rise to a secondary field being scattered outward from the object. The source of the secondaryfield is a movement of the electrical charges in the structure in response to the incidence fields impinging on theobject.

Figure: 4-23 A localized object interacts with an incident electromagnetic field. The field causes movement of elec-trical charge which gives rise a re-emitted (scattered) electromagnetic field.

Full analysis in even the simplest case in quantitative form is a very involved mathematical problem. The essential

Incident

FieldScattered

Field

+-

8/20/2019 GPR Annan 2003



80

physical behavior is discussed here and readers should refer to other texts on electromagnetics to get more details on

the subject.

For most analysis one can view the incidence field as a field which carries a certain amount of power per unit area PI.

The object acts like a mask which absorbs energy and casts a shadow behind it as depicted in Figure 4-24. In essence

the object can be treated optically to first order with the incidence signal power being extracted. This extracted power

may be absorbed internally into the object or it may give rise to re-radiated energy which makes it detectable at somedistance (the GPR goal).

Figure: 4-24 An incident field illuminates a target with a finite amount of power per unit area. The cross section ofthe target interrupts the power flow. If the target absorbs the power intercepted it will cast a shadow.

In general, one defines the power extracted from the incident field from the object as follows

(4-55)

where A is the cross sectional area for the target.

To first order, A is the physical area cross section presented to the field. In practice, however, the area is only a way of

representing the effect and is not always equal to the geometrical cross section of the object. Only in the optical limit

of infinitely short wavelengths will the geometrical and target cross section be equal. When the object is finite

compared to the wavelength then A is called the effective cross section, Ae.

The target may re-emit some or all of the power which it extracts from the incident field. Mathematically the re-

emitted signal, Ps, is expressed as

(4-56)

where indicates the fraction of the power which is reemitted . The combination of is called the

scattering cross section for the target.

Incident field withP power per unit

areaI cross sectional area

shadow zone

Pextracted AP I=

Ps ξAePI=

ξ 0 ξ 1≤ ≤( ) ξAe

8/20/2019 GPR Annan 2003



81

Figure: 4-25 Variation of the scattering cross section of a spherical target as a function of dimension normalizedagainst wavelength (Skolnik (1970)). While specific to a sphere, similar behaviour is displayed by any object of finitedimension. The response increases until the object is on the same order of size as the wavelength. For GPR to pene-trate through a heterogeneous material, it is highly advantageous that the GPR wavelength be large compared to the

scale of heterogeneity.

Perhaps the most informative study of a scattering object is the work by Mie described in Skolnik (1970) who studied

the behavior of a perfectly conducting sphere of arbitrary dimension for monochromatic (sinusoidal) excitation. The

basic result is depicted in Figure 4-25. Here the cross section is normalized to the geometric cross section

(4-57)

and is plotted against the normalized dimension

(4-58)

where the factor a is radius of the sphere and is the wavelength of the excitation signal.

The characteristics of the plotted cross section versus dimension are shown in Figure 4-25. When the dimensions of

the object are small compared to the excitation wavelength, the response is called the Rayleigh response. This

represents the response of a small object and the response varies as the fourth power of the excitation frequency.

When the object dimension approaches the same size as the excitation wavelength, the response is called the Mie or

the resonance response. Here effective area can exceed unity and will oscillate in amplitude. Physically the charges(currents) on the sphere have transit times which are comparable to the transit time or the period of the excitation

signal. As a result, a response enhancement through resonance or a response suppression when anti-resonance can

occur. As the dimension of the object gets larger, the normalized cross section approaches unity. At this point the

response is referred to as the optical response and the geometrical cross section essentially matches the scattering

cross section. The object is much larger than the wavelength and the sphere has a cross section equal to that of a disc

with the same diameter.

Ae N Ae

πa2

--------=

D N 2πa

λ---------=

λ

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



82

It is interesting to compare the Mie scattering response with the behavior of the thin layer reflection coefficient

discussed in section 4.8. The behavior is quite similar which is not surprising given that the same basic physics and

phenomena are involved.

Of much interest in GPR and many other physical problems is the case where the object is small compared with the

wavelength which is referred to as the Raleigh scattering regime. The strength of the scattered signal and hence the

scattering cross section very strongly with frequency. Typically the scattering cross section can be expressed as

(4-59)

where C is a constant, a is the radius of the object and f is the excitation frequency. One can see the scattering cross

section increases rapidly with frequency and slight changes in GPR frequency will give rise to much more intense

response from a localized object.

4.13 SCATTERING ATTENUATION

GPR signals are invariably transmitted through complicated media. The signals encounter heterogeneous electrical

and magnetic properties on many scales. GPR survey design requires EM wavelengths to be comparable in scale to

the objects to be detected. Smaller scale heterogeneities generate weak or undetectable responses but their presence

has an impact on the signals as they pass by. The heterogeneities extract some energy as the EM field passes and

scatter it in all directions and may absorb some though ohmic dissipation. The best although imperfect analogy is a

light bulb being viewed on a foggy night. On a clear night, the light may be visible at distances of many kilometers.

When fog (water droplets) are present, the light may only be visible at a few meters. While the water droplets are

likely larger than the light wavelength so the analogy is not perfect, the essence of the illustration is that the direct

signal is so disrupted on its path from the light to the observer that it never reaches the observer.

When GPR signals travel thought complicated non-uniform media, the direct signal is constantly losing energy. Its is

impossible to quantify the energy loss unless a specific model is used. Considerable research has been carried out on

this topic in other EM application fields under the general topic of “propagation in random media”. In this section, a

rough idea of the magnitude of the effective and how to quantify scattering attenuation is presented in a very

rudimentary fashion.

The GPR signal carries power through the medium. The Poyting vector as discussed in chapter 2 can used to quantify

power transfer. Figure 4-26 illustrates how scattering can be viewed from an energy viewpoint. At any point on the

wavefronts, the incident signal with power per unit area impinges on local small scale scatters which are

characterized by a spatial size, a, and number per unit volume, N.

Figure: 4-26 GPR signals are scattered by small heterogeneities in material properties which reduce the transmit-ted signal

ξAe Ca6f 4

=

Incident signal Transmitted signal

Signal scattered by small

heterogeneities

8/20/2019 GPR Annan 2003



83

If the power incident is P I, over a path length dL, the incident field will encounter N dL scatterers per unit area.

Mathematically, the transmitted signal PT is expressed as

(4-60)

where dP is the power extracted per unit area by the scatters in the path length dL. The details of explicitly computingdP can become very complicated. To keep the discussion simple, the scatters can be viewed as having a scattering

cross sectional area, A. Following a simple optical analogy, the area the scatter presents to the incident signal

indicates the power the scatterer extracts from the incident signal. The scattered power is then expressed as

(4-61)

which immediately means that the power will decline exponentially as the fields travel though the material. Re-

grouping one has the differential equation

(4-62)

which has the solution

(4-63)

where αsP is the power scattering attenuation coefficient.

Since the EM field power depends on the square of the electric or magnetic field, the fields will attenuate with a

scattering attenuation coefficient αs

=αs

P

/2. In other words, the electric (magnetic) field will decrease as

(4-64)

where

. (4-65)

To provide a quantitative measure of this and to stress what role frequency plays, the special case of very small

scatterers which can be treated as rayleigh scatterers is analyzed. As indicated in the previous section, scattering from

an object can be extremely complex. Even a simple spherical object (normally discussed as Mie scattering) ismathematically difficult.

The rayleigh scattering cross section can be expressed as

(4-66)

where C is a constant with units of m-4 Hz-4, a is the sphere radius and f is frequency. The critical point to note is that

dPPP IT −=

dLPA NdP I ⋅⋅⋅=

dLA-NlnPdP

dP

⋅⋅==

L-

0

LA-N

0 ePePP ⋅⋅⋅ ⋅=⋅= P S α

L-

0 eEE ⋅⋅= sα

2

A Ns

⋅=α

46 f aCA ⋅⋅=

8/20/2019 GPR Annan 2003



84

scattering cross sectional area is very strongly dependent on the scatterer’s radius and the excitation frequency. For

small metallic spheres, an approximate value for C is

(4-67)

where K h is the relative permittivity of the host medium.

To illustrate the magnitude of the attenuation, two cases are computed and tabulated. The first might represent

aggregate pebbles in concrete; the second stones in coarse till. (Note that the attenuation is expressed in dB/m which

is equal to 8.69 times the attenuation coefficient above which has the units m-1.)

Table: 4-2 Scattering Attenuation in dB/m

Case 1 Case 2

At the higher frequencies, the small object assumption becomes dubious but the essential points to note are that

scattering attenuation is strongly frequency dependent and it increases rapidly with frequency

One last point to note is that scattering attenuation must be added to the ohmic or material loss attenuation to

determine the full attenuation the GPR signal will see as it travels though a heterogeneous lossy dielectric medium!!

. (4-68)

Scattering losses will invariably be present. The current discussion is intended to provide an understanding of the

basic physics. The subject can dealt with in a lot more depth and still requires more study in the GPR context. The

effect of volume scattering was recognized very early by the radio echo sounding community (see Davis (1972),

Watts and England (1976)) as a limiting factor in temperate ice sounding. Its effect was more important than for most

soil, rock and NDT type GPR surveys because ohmic attenuation is very much smaller in ice. Scattering attenuation

quickly became the limiting factor in temperate glaciers where water pockets formed strong scattering centers and

radar sounding required reduction of the radar frequency to get deep penetration.

4.14 PROPAGATION DISPERSION

An important concept for GPR is that of propagation dispersion. As we saw in the previous section the attenuation of

GPR signals can be frequency dependent. When a pulse travels through electrically lossy ground material, both

attenuation and velocity can vary with frequency (see Annan, 1996).

A GPR pulse has a finite bandwidth over which energy is distributed as depicted in Figure 4-27. If the velocity and

attenuation vary with frequency, different spectral components will travel at different distances in the same time and

Frequency (MHz) N=1000/m3 and a=0.02m N=10/m3 and a=0.1m

1 1.6 x 10-13 1.6 x 10-9

10 1.6 x 10-9 1.6 x 10-5

100 1.6 x 10-5 1.6 x 10-1

1000 1.6 x 10-1 1.6 x 103

C 2 106 –

K h2⋅ ⋅= m

4 – Hz

4 –

scattering ohmictotal α α α +=

8/20/2019 GPR Annan 2003



85

will have different changes in amplitude when they travel over the same distance.

Figure: 4-27 A GPR pulse contains a broad range of frequencies. If velocity and attenuation vary with frequency,the pulse will change its shape as it travels.

Quantitatively one could view the attenuation and velocity as functions of frequency which vary slowly in many

situations. If one makes the assumption that the radar pulse spectrum contained is between frequencies and ,

one can understand the effect of propagation dispersion by the following approximate description of attenuation

velocity (essentially a local Tayler expansion).

(4-69)

(4-70)

In all physical situations velocity and attenuation will increase with frequency. Further for the vast majority of GPR

problems, changes in attenuation are far more significant than changes in the velocity.

The effect of changing attenuation or velocity is to have higher frequency signals travel faster but suffer more

attenuation. This causes a pulse which appears to travel somewhat slower and contains lower frequency content as it

moves away from the source. Figure 4-28 shows how the typical GPR pulse can change shape with distance.

ω1 ω2

α α ω1( ) ∆α ω ω1 – ( )+≈

v v ω1( ) ∆v ω ω1 – ( )+=

8/20/2019 GPR Annan 2003



86

(a)

(b)

Figure: 4-28 GPR pulses travel with little change in shape in low loss environments such as a). In higher loss mate-rials (b), the pulse decreases in amplitude rapidly and exhibits strong change in shape.

The impact can be viewed as filtering the pulse. If there is no dispersion, and are zero and the pulse willtravel with the velocity and reduce equally in amplitude at all frequencies as

(4-71)

The dispersion can be viewed as applying a filter to the pulse. The filter function will be

∆v ∆αv ω1( )

e α ω1( )x –

8/20/2019 GPR Annan 2003



87

(4-72)

The velocity change induces a phase advance relative to the constant velocity while the attenuation decreases

the high frequency amplitude. The effect on the spectrum of a pulse is sketched in Figure 4-29.

(a)

(b)

Figure: 4-29 The effect of frequency dependent velocity and attenuation changes the GPR spectral distribution. Inlow loss conditions (minimal frequency dependence) the spectral amplitudes reduce uniformly as in (b). In higher

loss conditions, low frequencies are less attenuated compared to high frequency spectral components as in case (a).

In the past, this translation of GPR signals to lower frequencies has been noted but seldom explained in terms which

provide the insight given above. Often practitioners have referred to the pulse as being shifted in frequency. The pulse

does not change its frequency, just that the relative amount energy versus frequency is changed. This is more

correctly called ‘colouring’ the spectrum; the spectrum with more red or longer wavelengths will penetrate further

through the material.

e

i ω∆v

v2

----------- ω ω1 – ( ) –

e ∆α ω ω1 – ( )∆x –

v ω1( )

8/20/2019 GPR Annan 2003



88

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 5-GPR Instrumentation

89

5 GPR INSTRUMENTATION

5.1 GPR MEASUREMENT OBJECTIVES

In order to put the specifics of instrumentation related to ground penetrating radar (GPR) systems into perspective, it

is important that we understand the measurement objectives. GPR as it is known today has developed as a tool for mapping the internal structure of materials, primarily the ground but not necessarily limited to that. In general we aredealing with lossy dielectric materials which quickly absorb the radio wave energy as the signals traverse through themedium.

The requirements for GPR instrumentation are vastly different from those required for airborne tracking and surveil-lance radars. About the only similarity is that we use the travel time of the signal in order to estimate a distance to thetarget. Unfortunately the term "radar" brings certain connotations to mind, particularly among lay people and also toelectrical engineers versed in the conventional uses of radar. As a result, clearly defining the measurement objectivesis required before the specific nature of the ground penetrating radar instrumentation needs can be understood. Thefollowing is an attempt to present how the instruments work in principle without overwhelming detail.

5.2 FUNDAMENTAL PHYSICS

Radar is an acronym for "Radio Detection and R anging". The acronym implies that we use radio waves which areemitted from a source to detect an object at a distance and determine the direction to the object as well as the distanceto the object. In order to detect an object, the object must re-emit some of the radio wave energy that impinges on it.This requires there be a change in the electrical properties from the surrounding host material. As discussed previ-ously, changes in dielectric permittivity and electrical conductivity cause scattering of radio waves (electromagneticenergy). By detecting this scattered energy, it is possible to detect and position the sources of the scattered energy.

Implicit here is the return signal and the fact that there is a return signal. The magnitude and character of the returnsignal are controlled by the geometry and the impedance contrast of the object generating the return signal. Unfortu-nately the return signal is not necessarily unique and trying to uniquely identify the target on this basis alone is some-times impossible. Even advanced military radars have great difficulty identifying metal objects in a uniformenvironment. For the GPR application, the problem is orders of magnitude worse because of the highly variable hostmedium electrical property conditions that can be encountered.

Figure 5-1 illustrates the general concept of how this detection is carried out. We have an object with certain electrical properties immersed in another medium which has a different set of electrical properties. A source of electromag-netic energy, which transmits its signals in the form of radio waves, sends out energy towards the target and scatteredenergy is detected by a receiver.

Figure: 5-1 Concept of using reflected or scattered signal to detect and define an object. The distance to the objectand some information on it's composition can be obtained from reflected signal. Implicit in the full three dimensional positioning is the ability to sense or detect the signals in or from specific directions or make measurements in a num-

ber of different positions.

8/20/2019 GPR Annan 2003


5-GPR Instrumentation GPR Principles, Procedures & Applications

90

The physical items and procedures involved are as follows.

a. An electrical circuit is required to generate time fluctuating voltages in the desired frequency range insome controlled and monitorable manner is required.

b. An efficient means of translating this electrical signal into a electromagnetic wave field which cantraverse (without cabling or connections) through the host medium to the target is required. The trans-

ducer which does this is normally referred to as an antenna. An antenna transforms electrical voltage sig-nals into outward propagating electromagnetic wave energy and vice versa. The source(s) and antennaare normally referred to as the transmitter.

c. The target must generate scattered energy. As discussed in a previous chapter, this requires a contrast inelectrical properties and suitable geometrical sizes compared to the wavelength and distance from thesource such that some detectable signal can be returned to a measurement device.

d. A detection system (receiver) normally consisting of an antenna and electronic circuitry which can detectelectromagnetic waves and transform this information into an electrical voltage or current which can berecorded and viewed.

The fundamental physical behavior is the propagation delay between the time that the source emits its signal and thetime at which any echoes return back to the detector. This time delay is determined by the distance to and from thetarget divided by the speed with which the waves propagate through the host material. The essence of GPR (and allradars) is to measure this time delay. The larger the time delay the greater the distance to the target, assuming uni-form velocity conditions.

No matter how the radar data are acquired, the most common display is one of showing the signal amplitude versustime. Figure 5-2 illustrates the basic concept. The data display is one which shows the amplitude of the return signalversus delay time. Detection presupposes the return signal from the target is bigger than background signals. Whilenot usually stated, this detected signal contains responses from a range of spatial directions as well as differing dis-tances. The real world is three dimensional - not one dimensional.

Figure: 5-2 Typical data recorded by a GPR for analysis is an amplitude versus delay time. Events from threedimensional space are mapped into a one dimensional time record.

The preceding discussion addresses the measurement of distance to a target. In addition to knowing the distance of the target, it is also important to know the direction to that target. Two approaches are used to measure direction asdepicted in Figure 5-3. When the direction in which the source emits its energy and the direction from which energyis detected with the receiver are controllable, a beam steering process is used to angularly position the target. In other words, the antennas are pointed at the target and antenna orientation is measured.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



91

a)

b)

Figure: 5-3 a) Source receiver signal directionality used to obtain angular position information. b) Multiplicity ofmeasurement positions used to define position.

If the source and detector directivity are low or non-existent, then moveable source and detector or multiple sourceand detectors can be used to locate a target by triangulation. From the basic physics, what we need in instrumentationis as follows.

a. a means to generate the signal;

b. a means to emit the signal in the form of radio or EM waves;

c. a means to detect and measure the delay time of these return signals;

d. a means to record and display detected signal and delay time;

e. a means for controlled movement or aiming of the signal sources.

8/20/2019 GPR Annan 2003



92

5.3 GPR TIMING AND R ESOLUTION

The next step in instrumentation assessment is to define the timing and the frequency bandwidth requirements for GPR instrumentation. In general terms, GPR is used for looking at objects which can be as shallow as one centimeter and as far off as hundreds of meters. If we choose 0.01 to 100 meters as extremes for typical GPR exploration, thenthese provide an indication of frequencies, times and resolutions that are important in instrumentation.

Obviously there may be exceptions to these minimum and maximum depths of exploration, but these can be handled by the same type of analysis as we are discussing here. What we are trying to do is get a typical range of values for the parameters which have to be considered.

5.3.1 TIMING

If we take the minimum and maximum distances and combine these with the minimum and maximum velocities typ-ically encountered in most geologic or man-made materials explored by GPR, then we have minimum and maximumtime ranges for the objects that we may wish to detect. Table 5-1 summarizes this information and we can see that theextremes for the travel time of our signals will be somewhere between 67 picoseconds and 6,700 nanoseconds. Theseare very small times on an every day scale.

Table: 5-1 Computation of Maximum & Minimum Delay Times

Delay time =

Maximum time =

Minimum time =

Maximum distance 100 m

Minimum distance .01 mMaximum velocity 0.3 m/ns (air)Minimum velocity 0.033 m/ns (water)

Maximum time = 6060 nsMinimum time = 67 ps

5.3.2 R ESOLUTION AND BANDWIDTH

In addition to the travel times we also have to consider how we are going to discriminate whether or not more thanone target is present. This aspect of the problem leads to the concept of system bandwidth. Discrimination of morethan one target is summarized in Figure 5-4. If two events occur on a record then they have to be sufficiently sepa-

rated in time so that the two events are clearly seen as distinct entities rather than one larger event (Berkhout (1984),Knapp (1991), Kallweitt & Wood (1982)).

Resolution is most easily understood by considering two pulses which occur as amplitude oscillations versus time.The pulse envelope, as depicted in Figure 5-4, is the dotted line which encloses the oscillatory radar pulse. If thetravel time to two individual targets is similar the pulses (and the envelopes) overlap. It is generally accepted that thetwo events can be distinguished as opposed to having one large event if the targets are separated in time by a time dif-ference of half the envelope width.

2 Dis cetan×velocity

--------------------------------

2 maximumdis cetan×minimumvelocity

----------------------------------------------------------

2 shortestdis cetan×maximumvelocity

----------------------------------------------------

8/20/2019 GPR Annan 2003



93

a)

b)

c)

Figure: 5-4 a) Two radar return events are well resolved if the two transient signals are clearly separated in time asdepicted here. The dotted lines, called the pulse envelope, are totally separated. b) When two events overlap each

other in time, the question arises as to whether one or two events are present. c) Examining pulse envelopes only, twoevents are said to be resolved when separated by a time w/2.

For GPR, there are two aspects of resolution which have to be considered. The first is referred to as "transmitter blanking", and the second is referred to as dual target separation and discrimination. The two concepts are illustratedin Figure 5-5.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



94

a)

b)

c)

Figure: 5-5 a) Transmitter blanking occurs when the direct signal travelling from the transmitter to the receiveroverlaps in time reflected signals. b) If two targets yield similar path lengths the differences in travel time can be

small causing reflected pulses to overlap. c) To resolve two events, the path length difference must exceed half the pulse width multiplied by the velocity.

Transmitter blanking is the phenomenon known as the inability of a receiver to detect signals until after the transmit-ter has finished transmitting. The transmitting source usually emits a very large signal and, if the receiver is in close proximity to transmitter as is usually the case, then the receiver will see the very large direct transmitted signal. If thissignal is sufficiently large that the receiver electronics are overloaded the receiver will not detect any reflected signalsuntil after the transmitter has ceased emitting its signal.

The second aspect of resolution is separating responses from two closely spaced scattering targets. In this case theseparation in time is proportional to the difference in distances to the targets divided by the propagation velocity inthe medium.

∆Lengthw

2---- velocity×>

8/20/2019 GPR Annan 2003



95

This discussion supposes that the radar signal has been processed to make W as small as possible. The envelopewidth in time, W, is obviously important and measures the spatial resolving capability of the radar.

The envelope width is usually the half width of the envelope. (i.e. it is the time between the points on the envelopewhere the envelope is greater than 2 of its peak amplitude or 6 dB points). Sometimes other definitions are used suchas -3 dB or -10 dB points.

Considerable time has been expended here in discussing resolution because resolution is a very critical aspect of sys-

tem design. The resolution needs of a system translate directly into the frequency content or bandwidth required inthe transmitter and the detection system.

In practical GPR applications resolution is dictated by the depth of exploration. Realistically one does not require thesame resolution needed for targets at one centimeter as for targets at depths of a hundred meters. A rough guide is thatthe resolution should be on the order of the maximum depth of exploration divided by 100.

(5-1)

The above relationship is reasonable representative of modern commercial systems and shallow sounding applica-tions. While for very shallow depths the 100 factor is appropriate, for larger depths the resolution available with mod-ern systems can approach 1/1,000 of the depth of exploration.

Taking this basic analysis, we can then say that the time between events is roughly given by the maximum depthdivided by the minimum velocity divided by 100.

(5-2)

From this analysis we have guidance as to what values of W (pulse width) are likely to be useful. The results are sum-marized in Table 5-2.

Table: 5-2 Summary of depth, resolution, and bandwidth typically needed in GPR systems

Bandwidth =

*Note: A material with a permittivity of 9 is assumed in calculated envelope width and bandwidth.

5.4 GPR BANDWIDTH

Instrument bandwidth determines a systems minimum response time (for a GPR this translates into resolution). Band-width and response time (i.e. radar envelope width) are inversely depicted in Figure 5-6.

Maximum

Depth (m)

Resolution

Required(m)

Envelope

Widthin (ns)

Required

Bandwidth(MHz)

0.1 0.001 0.02 50000

1.0 0.010 0.20 5000

10.0 0.100 2.00 500

100.0 1.000 20.00 50

∆r dmax

100-----------≈

∆t2

vmin

----------dmax

100-----------×≈

2

1---

1

Pulsewidth-----------------------------

8/20/2019 GPR Annan 2003



96

At one extreme, a narrow bandwidth amplitude modulated sine wave has a very oscillatory pulse which has a finitetime duration. In the frequency domain, the frequency bandwidth of the signal is quite small. In radar jargon, this isreferred to as a pulsed CW (continuous wave) signal.

a)

b)

Figure: 5-6 a) A narrow bandwidth signal is a long oscillary pulse characterized by single dominant frequency, f c.

b) A wide band signal has few if any oscillations, a short time duration and broad frequency content which can at bestbe said to be centred at a frequency, f c .

The other extreme is the ultra wide band monocycle pulse. In this case, a single cycle of a sine wave is the signal. The bandwidth and the centre frequency are about equal.

GPR pulses are characterized by the bandwidth to centre frequency ratio

(5-3)

Every effort is made to make R as large as possible for GPR. The goal is always to maximize B and minimize f c . The practical limiting value of R is about unity.

A surge of interest in impulsive source devices has created a great deal of interest in ultra wide bandwidth technology.A more general form of R is

(5-4)

where f max and f min are the maximum and minimum cutoff frequencies in the spectrum. The definition avoids having tospecifically identify centre frequency, f c.

R B

f c----=

R 2 f max f min – ( )

f max f min+----------------------------------=

8/20/2019 GPR Annan 2003



97

When constructing radio frequency instruments and antennas, keeping R small has been desirable because it makesthe job easier. If a desired resolution is set, then R can be kept small by increasing f c. Unfortunately this is not desir-able for GPR applications.

The reason for this is the high attenuation radio waves encountered in soils, rocks and man-made materials. In gen-eral, attenuation increases as frequency increases. The attenuation in natural materials is a combination of electricallosses and considerable scattering loss which both decrease with decreasing frequency. Since attenuation can be very

severe, the lower the frequency the more likelihood of obtaining signal penetration into the medium. Figure 5-7shows the general characteristic of attenuation versus frequency typical for lossy materials. As a result, GPR systemsattempt to operate with as low a centre frequency as possible. The result has been that most GPR's try to achieve amonopulse type character (i.e. R=1) and are most appropriately referred to as ultra wide band radars or impulseradars.

Figure: 5-7 Attenuation of radio frequency signals in soils, rocks and man-made materials always increases with frequency, forcing GPR's to operate at as low a frequency as practical given bandwidth requirements.

There are obviously some situations where the carrier frequency shouldn't be taken too low. If the bandwidth require-ments suggest that the carrier frequency should be below the transition frequency of the host medium (see chapters 2& 3), then one will be operating in a diffusive environment for the EM signals rather than a propagating environment(Annan, 1996). In this situation, one should be more careful about system selection and design.

5.5 SYSTEM ELEMENTS & CHARACTERIZATION

A GPR system is conceptually very simple. A block diagram depicting a GPR system is shown in Figure 5-8. Theheart of the system is the timing unit which controls the generation of the radar signal and then the detection of receive signals as a function of time.

Figure: 5-8 Block diagram depicting main components of a GPR system.

8/20/2019 GPR Annan 2003



98

Characterization of a radar system is an extremely complex task. A simple block diagram of a GPR system is shownin Figure 5-8. There are many elements in the system which, although not apparent on the surface, impact the opera-tion and use of the system. The major factors which govern GPR classification and characterization are summarizedin Figure 5-9. The following is a brief summary of the impact of each of these items.

SYSTEM CHARACTERISTICS

•Signal capture method

•Signal Processing

•Performance factor

•Centre Frequency and Bandwidth

•Antenna Patterns

Figure: 5-9 Characteristics which define a ground penetrating radar system.

5.6 GPR SIGNAL ACQUISITION

Few users of GPR delve into the details of GPR signal acquisition and real time processing. Numerous issues at thetechnical core of GPR systems can generate misleading concepts. The ideal GPR system is shown in Figure 5-10. Inthis system, the transmitter electronics deliver an electronic signal to a transmitting antenna which energizes the sur-roundings. The receiving antenna detects the transmitted fields and returns an electrical signal.

Figure: 5-10 Ideal GPR system with full capability and digital output.

In this perfect system, the output signal from the transmitter electronics, p(t), being fed to the transmitting antennaand the signal returned from the receiving antenna, r(t), would be digitized at a rapid rate through a high dynamicrange analog-to-digital (A/D) converters and recorded, processed and displayed.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



99

Unfortunately this ideal system while practical for audio frequency signals and the basis of modern seismic systems isimpractical at the frequencies of GPR systems. The reason for this is that A/D converters having sufficiently highsampling rates with sufficient dynamic range do not exist. Specialized transient samplers with 500 to 1000 MHz con-version rates have been implemented (Wright et al, 1994) but with high power demands and limited dynamic rangeunless stacking is employed. Alternate techniques, which have been used for many years, are still the mainstay of signal acquisition in GPR systems. The objective in these designs is to measure the ground response without havingto directly digitize or record the high radio frequency signals. A combination of repetitive transmissions and special-ized analog circuitry are used. The following is a very abbreviated discussion of the approaches employed.

5.6.1 CORRELATION ACQUISITION GPR

One technique for creating a lower signal rate GPR acquisition is to use a correlation based signal detection process.This approach is depicted in Figure 5-11. In such a system, the transmitter electronics generate a wideband (usuallyrandom) signal. This signal could be noise or a digitally generated pseudo-random binary sequence.

Figure: 5-11 Simple correlation based GPR system.

The detector is a correlator which is an analog signal multiplier and integrator. In this case, the correlator multiplies areference signal (i.e. the transmitter drive signal) and the input signal together and integrates the output. The result isa low frequency (essentially DC) signal which can be sampled with a high dynamic range, low speed A/D converter and then recorded.

In the simple system shown here, the control system varies the delay of the transmitter signal with respect to thereceived signal. This delay, which is the GPR time delay, causes the correlator output to change. A good summary of an implementation of such a system is described by Wills (1992).

The sampled correlator output versus delay is the impulse response of the ground convolved with the antenna charac-teristics and the auto correlation of the excitation (transmitter drive) signal. The benefit of this style of system is lowspeed signals and simplicity. Data acquisition speeds can be too slow for many applications.

5.6.2 FREQUENCY DOMAIN MIXER GPR

Another technique for acquiring GPR high frequency signals without having to deal with the high frequency signalsdirectly is depicted in Figure 5-12. In this style of system a mixer is used. A mixer is very similar to a correlator.Another term for mixing signals is heterodyning.

8/20/2019 GPR Annan 2003



100

Figure: 5-12 Frequency domain mixer approach to GPR system.

A mixer is an analog circuit designed to multiply sinusiodally varying signals. As depicted in Figure 5-13, a mixer combined with a low pass filter outputs a resulting signal which varies sinusoidally at the frequency difference

between the inputs. If the two signals have the same frequency, the output is a DC voltage proportional to the ampli-tude of the input signals. This low frequency (DC) voltage can be sampled with a low speed high dynamic range A/D converter and recorded.

Figure: 5-13 Mixer is a circuit which multiplies sinusoidally varying signals

In the frequency domain approach, the frequency output of the transmitter electronics (which is a sinusoidal) is con-

trolled. The output of the transmitter electronics is mixed with the input from the receiving antenna. By filtering the

transmitter signal through a 90o phase shift it is possible to measure the received signals that are both in-phase with

the transmit. This technique (which has been used in the majority of commercial GPR systems to date) is called the

equivalent time sampling (ETS) approach. Figure 5-14 depicts an equivalent time system.

By appropriately changing the oscillation frequency, one can measure the ground transfer function multiplied by the

antenna characteristics over a frequency range. This transfer function can then be fourier transformed to provide the

transient response of the earth. The transient response is the impulse response of the ground convolved with the

antenna impulse response. Good examples and discussions of this style of GPR system implementation are given by

Stickley et al (1998) and Noon et al (1994)

8/20/2019 GPR Annan 2003



101

5.6.3 EQUIVALENT TIME SAMPLING GPR

This technique (which has been used in the majority of commercial GPR systems to date) is called the equivalent timesampling (ETS) approach. Figure 5-14 depicts an equivalent time system.

Figure: 5-14 Time domain approach to a GPR system.

In an ETS system design, the transmitter generates a controlled output pulse with a specified temporal shape (and fre-quency content). The excitation pulse is fed to the transmitting antenna and to a signal acquisition circuit (if sodesired).

The sampler is very fast switch. The switch is linked to the transmitter control. The transmitter control clock isdelayed by a time n dT and used to toggle the sampler.

The sampler, shown in Figure 5-15, is a high-speed analog switch which closes for a very short duration. When theswitch is closed, the input voltage charges a capacitor. The resulting output voltage is proportional to the integral of

the input signal over the time interval of switch closure. If the time window is very short compared to the rate of change of the input signal, then the output voltage is a measure of the input voltage at the time delay n dT.

Figure: 5-15 Sampler is just a fast opening and closing switch.

8/20/2019 GPR Annan 2003



102

The equivalent time sampling approach uses the transmitter control and an adjustable delay time to sample successive points along a transient wave form. The Vout result is a slow varying replica of the high frequency signal which can besampled by a low speed analog converter. This process is the electronic equivalent of the stroboscope concept whichis used to slow down or freeze repetitive motion. Original discussion of ETS signal capture is given by Hansen(1942). ETS has been part of high frequency signal capture oscilloscopes for decades and has advanced to a veryhigh level with constant improvement and evolution.

In the majority of time domain GPR systems, the transmitter electronics and antennas are designed to be as invariantas possible. The objective is to make these components stable so that a second sampler to record the transmitter output signal is not required.

The resultant signal r (n dT) is the sampled version of the transient response of the transmitter output convolved withthe impulse response of the ground and the impulse response of the antennas. In the ideal system, the transmitter electronics would output a perfect function (spike) and the received signal would be the impulse response of theground convolved with the antenna system impulse response.

5.6.4 COMMON SIGNAL CAPTURE ISSUES

The preceding discussion provides a simple conceptual view of how GPR’s acquire signals. When faced with the dis-cussion of electronic implementation, the average user can quickly be overwhelmed by electrical engineering jargon.

Generally, commercial radars have been of the ETS type. These radars have often been referred to as impulse or baseand radars because the driving signal is usually an impulse.

Recent electronics developments have made sinusoidal synthesizers and mixers more robust and inexpensive. Thishas lead to researchers implementing step frequency systems. The step frequency systems are based on the frequencydomain approach depicted in Figure 5-12.

In an ideal world, all of the systems presented will generate equivalent results. The differences lie in the electronicembodiment of the circuitry. Issues such as the degree of linearity in the analog components, component stability andcircuit complexity all affect which approach may be better than the other in a particular application.

To date, the time domain systems have been preferred because the raw signal acquired is representative of the groundresponse with no (or minimal) signal processing required. As computers in small packages become available, stepfrequency systems using the frequency domain approach have started to appear. Computer implemented fast fourier transforms can generate transient output in these systems. In general, the ETS time domain approach is more tolerant

of non-linearity in system elements.

There are some common fundamental issues which have not been addressed in the discussions so far.

a. The antennas of a GPR system need to be close to the ground so that good ground coupling will occur

and emissions into the air are minimized.

b. When GPR antennas are close to the ground there is a considerable degree of variation in the ground

impedance which has an impact on the transfer function (impulse response) of the antennas. As a result,

the antenna response can not be assumed as invariant but will vary from place to place. The measure-

ments in all the GPR systems are contaminated to some degree because the transfer function of the

antennas is a variable which is difficult to control and even more difficult to measure.

c. Modern GPR system designs have placed major emphasis on minimizing the antenna variability withground impedance conditions. The quality of a GPR system is directly proportional to how well the

engineering of the antenna (and associated electronics) minimizes this variation.

d. Antennas that are mismatched or vary with ground conditions will generally exhibit reverberations

(commonly called “ringing”). Instead of an ideal impulse being emitted or synthesized post acquisition,

the antennas will generate a repetitive train of emissions that are spread out in time. This reverberation

masks the ground response and makes it difficult to interpret the GPR response.

δ

8/20/2019 GPR Annan 2003



103

e. In systems where mismatch is common, users resort to “background” subtraction for reverberation

removal to see the raw ground response. Background subtraction measures the primary or direct signal

and attempts to subtract or filter it out in some manner. Systems that depend on this process suffer from

the fact that a part of the ground response will often be removed in the process.

5.7 R EAL-TIME SIGNAL PROCESSING FUNDAMENTALS

The simplest model for the received signal is the transmitter electronics output convolved with the antenna systemimpulse response, a(t), and the impulse response of the ground, g(t). Note that this simple model assumes that theantenna directivity is independent of frequency which is definitely not true for most antennas!! Despite this limitingreality, the model is reasonable valid for most GPRs which use small dipolar antennas.

(5-5)

Note that ** means convolution in this discussion.

When a correlating GPR architecture is used, p(t) represents the auto correlation function of the excitation signal. Inthe frequency domain, obtained by fourier transform, the convolution operator becomes a simple multiplication of thetransfer functions.

(5-6)

Note that ω represents angular frequency. Capital letters indicates transfer functions while lower case letters repre-sent impulse response functions.

In a frequency domain style system architecture, one has

(5-7)

where I and Q are the in-phase and quadrature part of the received signal.

5.7.1 DECONVOLUTION

In GPR, the obvious measurement goal is to recover the impulse response or transfer function of the ground (g (t) or G ( )). A scheme is needed to remove the antenna response and transmitter (source) excitation characteristics toleave the pure ground response. The extraction of g(t) is referred to as “deconvolving out” the measurement systemresponse. This is never possible in the pure GPR context since nulls occur in the antenna transfer function at 0 fre-quency and possibly at other frequencies. Similarly, the transmitter output may have nulls in its spectrum .

Ignoring these very real issues for now, the ground impulse response can be mathematically expressed as

(5-8)

where IFT stands for inverse fourier transform. The fact that A(ω) and P(ω) have nulls represents a numerical prob-lem that must be addressed in real systems. At the spectral nulls, the ground response is undefined and the computedresult will be determined by the relative character of the error (noise) in the measurements of R, A and P. Blindly car-rying out the above “deconvolution” usually generates a result which is random garbage signal bearing little or norelationship to the true ground response.

In the early days of GPR, attempting such a process was not realistic given the limited and analog-only processingcapability available. With the advent of powerful microprocessors and DSP devices, such processing can be imple-mented in real-time. Since GPR has been a useful device for many years, it is useful to review the compromise solu-tions which give useful results.

p(t)a(t)g(t)r(t)

)ωP()ωA()ωG()ωR( ⋅

)ωQ(i)ωI()ωR( ⋅

ω

⋅ )ωP()ωA(

) R(ωIFTg(t)

8/20/2019 GPR Annan 2003



104

The most obvious solution is make the combined antenna and excitation signal mimic a perfect impulse ( functionmathematically). The resulting devices are called impulse or baseband radars in radar jargon. Now-a-days suchradars are called ultra wideband (UWB) radars.

When this occurs, then

(5-9)

Physical reality limits how well this can be achieved as the discussions of the nulls in the antenna transfer function previously suggest.

By suitable electronic design and antenna construction, the system response can be reduced to a short time-durationwavelet, w(t). The received signal then has the form

(5-10)

For the most part, this is the approach taken by GPRs today. Engineering design to make w(t) a very compact timeduration signal is the prime focus of system developers. All current practical GPRs use r(t) (or possibly someenhanced form thereof) and not g(t). For many applications this is not a great limitation and very useful results areobtained.

In conventional radar, the need to make w(t) compact is not so critical and can often be achieved by increasing theoperating frequency while maintaining a constant bandwidth. GPR always needs the maximum bandwidth at the low-est possible frequency.

Considerable effort has been devoted in recent years to post-processing of GPR to deconvolve the data by a variety of processing techniques (Turner (1992), Majala (1992)). In general these attempts have produced marginal benefit for data obtained with well engineered systems where w(t) has been refined to be very compact by the system hardwaredesigners.

There are many ways of carrying out deconvolution (weiner, least squares, parametric, spectral whitening, etc.) in amore sophisticated manner than that suggested mathematically above. The seismic processing industry has refinedthese concepts to a very high level. An excellent overview of the seismic advances and techniques which also applyto GPR is given by Yilmaz (1987). Simple stabilization of the above operation can be achieved (in Equation 5-11) byadding a whiting contribution to the denominator which can be adjusted to make the deconvolved output

(5-11)

By adjusting empirically β , the output can be shaped. In some instances improvements in estimating g(t) can beachieved. Unstated here is the need to measure p(t) and a(t) or A(ω) and P(ω) which is not a trivial matter.

Since most GPRs use a form of r(t) rather than attempting to get to g(t). The form of w(t) is of considerable interest.There is no specific form that fully represents the full antenna behavior but a good approximate model for modellingand simulation is as follows.

5.7.2 GPR WAVELET MODEL

The radiated wavelet from a GPR system is a complicated function of the antenna construction and the electronicsdrive circuitry. For impulse style ultra wideband systems a simple mathematical model is available for helping withnumerical simulation.

The form of signal is

(5-12)

δ

g(t)r(t) =

w(t)g(t)r(t)

β)ωP()ωA(

) R(ωIFTg(t)

T)q)v(t(1)2

Tq)v(t(2v(t)T)q,s(t,

8/20/2019 GPR Annan 2003



105

where 0 < q < 1 (damping factor)

q – damping factor (0 to 1)

f c – center frequency

and

(5-13)

and

(5-14)

The damping factor q depends on system design. As q varies from 1 to 0 the wavelet (t) waries between one cycleand one and a half cycle shape as shown in Figure 5-16. The amplitude spectra for various values of q are shown inFigure 5-17.

Figure: 5-16 Typical GPR model wavelets for a range of damping parameters.

cf

1

7

q)(1

3

2T

−

otherwise0

Tt0

2

T2

Tt

cosπ12

1v(t)

=

<

−

ω

8/20/2019 GPR Annan 2003



106

Figure: 5-17 Amplitude spectra of GPR pulses for varying damping factors

5.8 SYSTEM PERFORMANCE FACTOR

System performance factor, Q, is a measure of the ratio (usually expressed in decibels) of the source power to thereceiver noise power. The basic ideas are summarized in Figure 5-18. The reason the system performance factor isso important, is that it is the primary factor which determines the range to which the radar can see and the size of objects the radar can detect.

System Performance Factor

Q = 10 log10 dB

Q = 20 log10 dB

Figure: 5-18 The system performance factor Q provides a readily measured GPR parameter.

power noiseReceiver

power Source

voltageReceiver

voltager Transmitte

8/20/2019 GPR Annan 2003



107

In order to understand the impact of the system performance factor, it is best to refer to the radar range equation. Fig-ure 5-19 gives a block diagram of the radar range equation. It shows the sequence of events from the point where thetransmitter electronics generate a signal, (i.e., referred to as the source power) through to the signal which the receiv-ing electronics present to the display unit, (i.e., the received power). In order for a feature to be detectable thereceived power must be in excess of the noise level in the receiver.

Figure: 5-19 Block diagram showing the major steps in the radar range equation.

Analyzing particular situations with the radar range equation allows one to anticipate the range to which a target can be detected. For a given system performance factor value, the sum of all the losses of energy along the various pathsthat the signal will travel must not exceed the system performance factor; otherwise, the target will not be detectable.Detailed implementation in a GPR context is described by Annan & Davis (1977) and Annan & Chua (1992).

5.9 SIGNAL AMPLITUDE & R ECORDING DYNAMIC R ANGE

While GPR data can be acquired in a number of different ways using different recording procedures, the data arealmost invariably reduced to an amplitude versus time output for evaluation and display. The response is in a formthat approximates the impulse response of the system and the ground.

Systems are never perfect. There are a number of factors in the electronics and the actual physical antenna structureand support systems which will invariably cause the impulse to be blurred in time. There are two major issues whichmust be addressed. First, the question arises as to how well does the system approximate an impulse response andwhat does this do to the recorded ground response. Second, how does the dynamic range of recording impact measur-ing the ground response.

5.9.1 CHARACTERIZING SYSTEM R ESPONSE

The system response should generally respond as shown in Figure 5-20. The transmit signal has an onset at 0 timeand lasts for a duration of W. In a perfect world, all system signals would vanish after W. The signal during the inter-val 0 to W represents the excitation impulse.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



108

Figure: 5-20 Idealized GPR impulse response envelope amplitude (log) versus time. The key parts of the responseare the ambient noise level A3 (before pulse and at late time), the peak signal A 1 and initial residual system response,

A2.

With real electronic components, responses will linger in time after the ideal signal has terminated. In addition, with

a GPR system, the actual emitted signals and detected signals are launched and retrieved using an antenna structure.The antenna structure has (intrinsic to its character) a finite time delay for currents to move around the structure toachieve efficient launch and detection of signal.

In addition to the antenna, the antennas have to be carried, mounted and interconnected using cables and other mechanical structures. These ancillary structures can also have electrical currents induced upon them which in turngenerate re-radiated fields that occur at a time delay.

The standard GPR approach is to denote the signal which appears after the transmit pulse should ideally have turnedoff as the residual system response. This system response normally will “ring-down” or decrease with time after thetransmit pulse and fall below the noise level.

Echoes from the ground will return during this “ring-down” period. If the residual signal from the system is larger than the ground response, then the ground response will be masked and not visible. In general, the ground responsewill be largest at an early time and get successively smaller as time increases. Quite often a mathematical model

which describes signal amplitude , G(d), from the ground versus depth, d, is useful for discussion.

(5-15)

(5-16)

(5-17)

In this model, the ground response amplitude for a target at depth d is re-expressed in terms of attenuation, α, and thevelocity, v. If one assumes that time is equivalent to depth as in Equation 5-16, the ground response amplitude versustime can be expressed in Equation 5-17. In other words, the ground response on average will decay inversely andexponentially with time and be dictated by the velocity and attenuation in the ground.

For the basic design and analysis, the GPR response (no ground present) can be characterized by a simple model. Thecharacter of the system amplitude response versus time as shown in Figure 5-20 is expressed as follows.

A = A3 t < 0 (5-18)

d

d

eGG(d) o

α=

v

2dt =

αvt/2

o e

vt

2GG(t) −=

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



109

A = A1 0 < t < W (5-19)

A = A2 – B (t – W) W < t < tn (5-20)

A = A3 t > tn (5-21)

where

(5-22)

and

A1 - peak signal during the transmit pulse (5-23)

A2 - residual signal amplitude at time W (5-24)

A3 - noise level of the measurement (5-25)

B - residual “ring-down” rate (normally in dB/ns) (5-26)

Figure 5-21 through Figure 5-25 depicts the ground response superimposed on a typical GPR system response. Herethe ground response initial amplitude and time delay rate (αv/2) from Equation 5-17 are shown superimposed on thesystem response. In the example shown in Figure 5-21, the ground response is larger than the system residualresponse and decays slower. This is the ideal situation and the system residual has a negligible impact on the mea-surement of ground response over all delay times.

Ground conditions can impact both the ground response as well as the system residual response. As a result, a widevariety of conditions can occur depending on ground conditions. Some of these behaviors are depicted in Figure 5-22through Figure 5-24. The system residual response can mask some or all of the ground response over some time peri-ods.

In some situations, the system residual response may be predictable; if it can be estimated and subtracted, the systemresidual response can be reduced with respect to the ground response and the ground response made visible such asdepicted in Figure 5-25.

Figure: 5-21 Case where signals from the ground are larger and decay more slowly than system “ring down”.Ground response is always visible.

WB

AAt 32n =

8/20/2019 GPR Annan 2003



110

Figure: 5-22 Case where system “ring down” is faster than the ground response fall off but initial system residualis larger than ground response. Ground response becomes detectable after t g .

Figure: 5-23 Case where ground response is smaller than system residual response resulting in no detection of ground response.

Figure: 5-24 Case where ground response initially exceeds system residual but ground response decays morequickly than system “ring down” rate leaving ground response detectable only between time t g1 and t g2.

8/20/2019 GPR Annan 2003



111

Figure: 5-25 I n some cases the system response might be static and subtracted from the combined signal to make ground response visible at all times. This can apply to Figure 5-21 through Figure 5-24.

Unfortunately, the residual response is not easily determined and is not always stable. In particular, the residualresponse is often associated with the antenna system and its support structure. The “ring-down” or residual electricalcurrents in the structure are modified by the ground conditions which in turn makes the “ring-down” character vari-

able with measurement location making removal difficult to impossible and a trial and error procedure at best.

The key parameters to characterize a system are A2 which is the initial “ring-down” amplitude after the transmit pulsehas terminated and B, the decay rate or “ring-down” rate. These are very dependent on system configuration andapplication needs. It is possible to have systems with an A2 value which is –60 dB to –80 dB below A 1 and exhibit“ring-down” rates of –4 dB/ns.

5.9.2 DYNAMIC R ANGE

The preceding has shown the general character of the GPR signal amplitude versus time. In the initial days of GPR,these data were then transferred on to a paper recording or recorded on to an analog tape recorder. These recordingdevices usually had limited dynamic range - typically much less than the 80 to 100 dB or more which the GPR datarequired for recording. Now most systems digitize the signal at some point along the path and record digital data. Asa result, dynamic range of recording is less limited than it used to be.

Figure 5-26 shows typical signal levels that one may encounter in a system. The general character of the amplitudeversus time response depicted in Figure 5-20 and Figure 5-21 through to Figure 5-25, is shown together with a binarydynamic range scale on the left typifying the digitization levels of a 16 bit A/D converter. In this particular example,we have chosen A1 to be 50,000 mV and A3 to be 10 µV. Maximum and minimum recording of the receiver must beable to bracket A1 and A3.

Figure: 5-26 The typical GPR amplitude envelope is plotted on the right against the recording levels of a 16 bit(bipolar) analog-to-digital converter that might be employed in a recording system.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



112

In the configuration in Figure 5-26 a 16 bit bipolar A/D converter is assumed. This choice gives the smallest signalresolved as 1.5 µV if the maximum recording value of the A/D is A1 = 50,000 µV. For this example, the noise level,A3, is greater than the least significant bit of the A/D converter.

For rigid system geometries where there is no variation in the position of the components, the value of A1 can be con-trolled and the design objectives met. For bistatic antennas with arbitrary antenna geometries and orientations as wellas possible changes in frequencies of the antennas, the value of A1 can be highly variable and it is virtually impossible

to guarantee that A1 will always remain within the dynamic range of the recording system. As a result, it is commonto see the peak signal during the transmit pulse limited (clipped) by the recording system.

This is not a major problem for many GPR applications as the information in the transmit pulse is not necessarily use-ful. Loss of signal during this time is common in time domain radars and is called transmitter blanking. In the situa-tions where signals from shallow depths are critical, systematic attention must be given to the dynamic range of therecording system and the geometry of the antenna configuration to achieve signals which always remain withindynamic range.

In some instances, signal dynamic range can be decreased by applying a judicious amount of time varying gain to theradio frequency signal prior to recording and digitization. Figure 5-27 through to Figure 5-31 depict the basic con-cepts here.

For the signal shown in Figure 5-27, the dynamic range of the signal present is greater than the dynamic range of therecording device indicated.

By applying a judicious amount of amplification to the signal versus time, such as depicted in Figure 5-28, thedynamic range of the signal can be compressed. The time gain in Figure 5-28 attenuates the signal during the trans-mit pulse and amplifies the signal at an every increasing amount with time.

If the time gain is appropriately selected, the resulting signal will be compressed into the dynamic range of therecording system such as depicted in Figure 5-29. If the time gain is too large, the signals will reach the limit of therecording system and be clipped off and flat lined, such as shown in Figure 5-30.

On the other hand, if the time gain is not sufficient, the signal levels may be too small to be recorded by the recordingsystem and there is no signal in some or all parts of the record as shown in Figure 5-31.

Figure: 5-27 Example where recording system dynamic range is smaller than that of the GPR signal.

8/20/2019 GPR Annan 2003



113

Figure: 5-28 Example of a time gain function that might be applied to the data in Figure 5-27 to compress dynamicrange. The plot shows the amplification that would be applied to the GPR signal.

Figure: 5-29 If the time gain is selected very carefully, the resulting data can be compressed into the recordingdynamic range. Note that this process may cause noise signals to appear as large or larger than real signals.

Figure: 5-30 If the time gain is too large, signals may be over amplified, exceed the recording dynamic range andlimited or dipped. (Dotted line indicates where data exceed reading dynamic range).

8/20/2019 GPR Annan 2003



114

Figure: 5-31 If the time gain is too small, signals may not be recorded because they are below the recording thresh-old. The dotted line shows where data are too small for recording to detect.

In systems which have used time gain, the adjustment has often been left to the user. Unfortunately what often turned

out to be a good time gain at one location was not appropriate at another location resulting in recording of no usefulor distorted amplitude data.

In theory it might be possible to make a system which dynamically adjusted at every location to maximize the data insuch a way to fit recording dynamic range. This is complicated and difficult to carry out although attempts have beenmade. While it looks simple, this kind of adjustment in the RF circuitry can introduce many artifacts and distortionswhich may severely degrade data quality.

A point to note for new GPR advocates is that some ETS (equivalent time sampling) GPRs have analog time gainwhich is applied to the analog sampled signal. This type of time gain does not address the dynamic range limits of theradio frequency electronics in the signal capture circuitry and is not of much benefit given modern A/D perfor-mances.

At present, modern general purpose systems do not use time gain in the radio frequency section of the receiver. Timegain is regularly applied after digitization and recording when displaying the data on screens or printing on paper.This dynamic range adjustment or time gain is reversible by the user and can be adapted interactively to meet user specific needs. Using post recording time gain is preferable to applying it in a irreversible manner before recording at present. Hardware time gain can be very useful and with electronic advances some variations of RF time gain willappear in general purpose systems.

5.9.3 SUMMARY

When attempting to understand the details of GPR system behavior, one needs to know the dynamic range of therecording system, the maximum signal likely attainable (A 1), the typical noise level (A3), the after pulse initial “ring-down” amplitude (A2) and the “ring-down” rate (B). When these parameters are known and understood, it becomes practical to understand ground responses and put them into the context of what is recorded. These basic parametersare key characteristics of a GPR system.

GPR records are frequently impacted by some or all of the amplitude artifacts and distortions described here. Astechnology matures and systems improve these artifacts are steadily being reduced. It is virtually impossible to elim-inate some of the effects that are controlled by ground conditions and application constraints which are beyond thesystem users and designers

8/20/2019 GPR Annan 2003



115

5.10 CENTRE FREQUENCY AND BANDWIDTH

Frequency and bandwidth go hand in hand in a GPR system. The operating frequency of a GPR system is usuallygiven by indicating the centre frequency of its operating band. The associated factor is the bandwidth or the fre-quency range over which the radar has available power for use in sounding the ground. The bandwidth and the centrefrequency of a radar system are determined by several components in the system, the primary ones being the anten-nas. A block diagram showing how the various system components determine the frequency of operation and band-

width is presented in Figure 5-32.

Figure: 5-32 Simplified signal flow through system elements depicting controlling factors in centre frequency andbandwidth.

The centre frequency of the radar system has a major impact on the depth of penetration of the radar system. The rea-son for this is that as the frequency decreases the signal attenuation in the medium decreases. In addition as the fre-quency of the radar system is lowered, the impact of clutter and the masking of desired signals by responses fromsmaller scale features is reduced.

Bandwidth is the other factor in the system. The wider the bandwidth the better the resolution of the radar. The con-cepts of bandwidth and pulse duration and spatial resolution are depicted in Figure 5-33.

8/20/2019 GPR Annan 2003



116

Figure: 5-33 Illustration of the relationship between spatial, temporal and spectral aspects of signal which controlresolution.

For GPR systems designed for soil and rock, the objective is to achieve a bandwidth to centre frequency ratio of about1. Some example calculations of resolution are shown for antennas of this type in Figure 5-34. Figure 5-35 showsthe resolution versus system bandwidth for two geological setting, one a rock the other a wet soil. It is important tonote that the resolution in the ground depends on the electrical properties of the ground. Figure 5-35 indicates thisquite explicitly.

Figure: 5-34 Tabulation of spatial resolution versus antenna frequency.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



117

Figure: 5-35 Spatial resolution versus bandwidth illustrating how material velocity changes spatial resolution.

5.11 ANTENNAS

Antennas form the main part of a GPR system. Unlike antennas for communications needs and pulsed CW style air launch requirements, where the modulation of the radio frequency signal carries the information, GPR antennas mustcreate and detect electromagnetic fields to extract the actual field characteristics. In other words, the measurementobjective is the field amplitude in space and time, not the superimposed information.

In a GPR system, the transmit antenna must translate the excitation voltage into a predictable temporal and spatial

distributed field. The receive antenna must detect the temporal variation of a vector component of the electromag-netic field created by the transmit antenna and the ground response.

The following are desired antenna characteristics.

a. The exact source and detection locations must be definable.

b. The transmitter and receiver responses must be time and space invariant.

c. The vector character of the field linking the source voltage and received voltage must be quantifiable.

In practical terms, these requirements are difficult to achieve. The reason is that efficient field generation and detec-tion requires the size of antennas to be such that field travel time across the antenna dimension be comparable to thetemporal rate of change of the exciting voltage or field. In frequency domain terminology, the antenna dimensionmust be similar to the wavelength of the signals.

For efficient operation, finite size antennas must be used and they have the following characteristics.

a. Field creation and detection occurs over a spatially (and temporally) distributed region. (In other words,source and detection points are imprecise).

b. Field transit time or wavelength in GPR applications depends on the host environment and is not invari-ant. (In other words, antenna response cannot be perfectly invariant).

8/20/2019 GPR Annan 2003



118

c. A spatially distributed antenna means less precise vector characterization of the response since isolationof response to a single vector component becomes geometrically difficult.

The antennas that have been preferred for GPR have been short electric dipoles (or small magnetic dipoles). By mak-ing these antennas as small as possible without totally eliminating efficiency, a fair degree of faithfulness to desired predictable and invariant behavior can be achieved.

While other antennas have been used for GPR, such as log periodic or log spiral antennas, these antennas criticallydepend on their size to achieve a desired directivity or response character. While useful for communications applica-tions, such antennas suffer from pulse dispersion (in space and time) and smeared (mixed) vector response, whichrender the specific GPR desired response much less quantifiable.

5.11.1 DIPOLE IMPULSE R ESPONSE OR TRANSFER FUNCTION

The use of short electric dipole antennas yields relatively invariant radiated pulses and antenna directivity. Antennasare small provided

(5-27)

where t is the time duration for the fields or excitation voltage to change, L is the maximum spatial dimension of the antenna and v is the radiowave velocity in the host medium. (When the antenna is on the ground this is both afunction of ground properties and antenna proximity to the ground). Another way of expressing this is

(5-28)

where is the wavelength of sinusoidal signal in the host medium.

When antennas are small, the emitted field is roughly proportion to L2 (Schmitt et al, 1966). There is a great incen-tive from a signal detection viewpoint to maximize antenna size.

In practical GPR design the required application bandwidth , B, is defined and the centre frequency is selected to be

f c = B (5-29)

And electronic excitation is optimized to create and detect signals in a bandwidth B centered at f c. The dipole antennalength is selected such that

(5-30)

With proper resistive loading, driving such an antenna with a bandwidth limited impulse will generate a clean emittedwavelet such as depicted in Figure 5-36. This is the antenna impulse response.

tv

L∆<

∆

λ<

L

λ

cf

c

4

1L<

8/20/2019 GPR Annan 2003



119

Figure: 5-36 A multiplicity of antennas with centre frequencies at binary multiples allow full coverage of a wide spectral range. The spectra and corresponding wavelet in time for 4 antennas with centre frequencies f 0 , 2f 0 , 4f 0 and

8f 0 are as displayed. The increase in time resolution associated with increase in bandwidth (i.e. B=f c ) is apparent.

The important point to note is that application resolution dictates bandwidth and hence maximum allowable antennasize (which limits efficiency). As a result one antenna size will not effectively satisfy all application needs. Deepgeological sounding GPRs use long antennas which give greater efficiency but lower resolution. High resolutionGPRs such as for concrete imaging use small antennas but have lower efficiency and give much less depth of penetra-tion.

5.11.2 ANTENNA DIRECTIVITY

The directional characteristics of a short dipole antenna are controlled by the ground. Understanding antenna direc-tivity helps greatly with understanding the source of GPR returns from within the ground. The analysis of this prob-lem is complex but the basic characteristic can be fairly well explained.

Before continuing a great length, it must be stressed that GPR amplitude information has to date been used in a rela-

tive context. As a result, exact mathematical forms for fields and measurements have not been needed. As the tech-nology advances, more demand for quantitative measurements are arising and further advances on quantifyingresponses which use some of the numerical solutions discussed in the section of GPR modeling will become morecritical.

This current discussion addresses the far field radiation directivity for a short electrical dipole source in an overviewfashion only. Background for this can be found in references Annan (1973), Annan et al (1975), Engatta (1983) andSmith (1999).

A short electrical dipole and the associated coordinate system are shown in Figure 5-37. When the dipole is in a uni-form media, the relative electric field amplitude at a large distance is donut shaped as depicted in Figure 5-38. Noenergy is radiated from the ends of the antenna while energy is radiated uniformly in the plane perpendicular to thedipole axis. Figure 5-39 shows the cross sections through the donut commonly referred to as the TE and TM patterns.

8/20/2019 GPR Annan 2003



120

Figure: 5-37 Geometry and coordinate system for an electric dipole antenna. When placed on the ground, the X-Y plane would represent the ground surface.

Figure: 5-38 3D presentation of pattern of a small electric dipole is donut-like in a uniform material. There is no

radiation off the ends of the dipole.

Figure: 5-39 Cross sections through the dipole pattern referred to as the TE (H plane) or TM (E plane) directivity patterns.

8/20/2019 GPR Annan 2003



121

When the dipole antenna is placed on the ground surface, the pattern changes drastically as shown in Figure 5-40.The change in directivity is caused by the refractive focusing associated with the impedance change at the air-groundinterface. Note that this pattern represents the far-field radiated component of the fields. Near the antennas the fieldsare not as sharply defined by peaks and nulls and there are secondary lateral and evanescent fields along the air-ground interface.

Figure: 5-40 When the dipole is on the ground surface, directivity is drastically altered and depends on ground per-mittivity. The TE and TM patterns shown here are for typical ground permittivity.

The peaks in the TE (H plane) pattern occur at the critical angle of the air-ground interface.

(5-31)

A subsurface null occurs in the TM (E plane) pattern in the critical angle direction.

The antenna directivity is obviously ground dependent. As the ground properties change, the antenna directivity willchange. Figure 5-41 shows a sequence of patterns as K g is carried from 3.2 to 80. As K g increases, the pattern in theground gets pointed more downward.

Figure: 5-41 When the ground permittivity changes, the patterns change. The TE pattern is shown for permittivitiesrange from ice (low) to water (high).

=

=

−

g

g

cK c

v 1sinsin

11θ

8/20/2019 GPR Annan 2003



122

Another practical consideration is the effect of antenna elevation off the ground surface. In real field situations, sur-face roughness and the need to transport antennas over the surface can limit close ground contact. Antenna elevationstrongly affects antenna directivity as shown in Figure 5-42, the directivity is frequency dependent because the heighteffective is dependent on the relative distance in wavelengths above the surface. In the time domain frequencydependence translates into additional events and/or a dispersed radiated pulse.

Figure: 5-42 Antenna elevation also impacts on directivity. The change in directivity is shown here as a function ofheight normalized against wavelength assumed pure sinusoidal excitation.

The reason that the signal into the ground at high angles (beyond the critical angle) is strongly attenuated is caused bythe energy being of evanescent character in the air. The energy has to “jump the gap” between the antenna and theground and the amplitude of the evanescent signal decreases exponential with normalized height.

In addition, signals above the ground are strengthened and the direct and reflected signals from the air-ground inter-face no longer cancel one another. See section 4.10 for wavefront character.

5.11.3 SHIELDING

What is antenna shielding? Figure 5-43 shows the conceptual idea of a shield. With GPR, the antenna is normally placed close to the air-ground interface. The shield is a “container” which encloses the antenna. The objective is toimprove the antenna’s performance by placing the shield over it.

Figure: 5-43 Concept of a GPR antenna shield. The shield encloses the antenna to minimize coupling with signalsin the air.

What is the purpose? Referring to Figure 5-44, signals can travel from a transmitter to a receiver along a number of paths. Shielding is used to achieve the following goals:

• maximize the energy on the path AA’ to and from the subsurface target;

• minimize the direct transmitter to receiver energy on path B;

• minimize the energy which escapes into the air as on path CC’;

• minimize external electromagnetic noise as indicated by signals D.

8/20/2019 GPR Annan 2003



123

Figure: 5-44 A GPR system emits and detects radio wave signals. There are many possible signals and paths. Theobjective is to maximize the target response and minimize others.

Given these laudable benefits, what are the drawbacks? Antenna shielding requires a structure that has an electro-magnetic response. If the structure is not designed properly or is damaged, its electromagnetic response can be large.In addition to the signals shown in Figure 5-44, energy goes from the transmitting antenna to both the transmitter shield and the receiving antenna shield and then to the receiving antenna as indicated in Figure 5-45. The shield gen-erated signals can be large and reverberate for a long period of time.

Figure: 5-45 Antenna shields must interact with radio waves to be effective. The shield can generate additionalresponse which may be detrimental and interfere with the desired measurement unless extreme care is taken in the

shield design.

Besides the electromagnetic response of the shield itself, an effective shield has to be larger than the antenna. Thisleads to considerable transducer size, weight and manufacturing costs penalties. In a nutshell, attempting to shield anantenna can create as many problems as it benefits.

A series of radar traces are shown in Figure 5-46. (a) is the ideal response without external noise. (b) is the responseobserved when external noise is present. (c) shows the behavior which can occur if improper shielding is used. Thereverberatory signal sunning down the trace for along period of time is the transient response of the shielding. (d)

shows the result that a practical shield can achieve.

Shielded antennas can be readily constructed for higher frequency GPR systems, typically in the 100 MHz frequencyrange and above. The shields are about the same size as the antenna and use absorbing material to damp out theundesired signals.

At lower frequencies, practical size and portability dictates minimal or no use of shielding.

8/20/2019 GPR Annan 2003



124

(a) Unshielded noise free response

(b) Typical response with external noise

(c) Transient response when shielding is improperly implemented

(d) Results that can be expected with a practical well implemented shield.

Figure: 5-46 Typical GPR traces for various shielding configurations.

8/20/2019 GPR Annan 2003



125

Tips for GPR users are as follows.

a. Shielding is never perfect no matter what some may claim.

b. Question whether your application requires shielding. The highest fidelity and maximum depth of pene-tration at open sites may be obtained when shields are not used.

c. Always look for GPR events attributable to spurious signals. Even with the most ideal shield, spurioussignal leakage can and will occur.

8/20/2019 GPR Annan 2003



126

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 6-Modelling of GPR Responses

127

6 MODELLING OF GPR R ESPONSES

6.1 THE PURPOSE OF MODELLING

Modelling is critical to all geophysical methods. One can simulate the system behavior from excitation through to the

response which would be observed or measured. Modelling or simulation enables quantitative predictions of responses leading to both a better fundamental understanding and a sound basis for interpretation.

GPR modelling is critical for the following reasons:

• understanding of physical behavior and quantifying response;

• providing performance requirements for design of measuring instruments;

• predicting response and sensitivity to parameter changes;

• optimizing survey design;

• understanding how to process data to extract information;

• enabling interpretation at a variety of levels of complexity; and

• facilitating mathematical inversion and quantification of interpretation uniqueness.

In summary, modelling underpins translation of geophysical observations into useful information (knowledge). GPR is no exception.

6.2 GPR MODELLING TYPES & DEFINITIONS

Modelling GPR responses has seen considerable evolution. Prior sections of these notes have been derived from fun-damental physical principles that are a key aspect of modelling. Modelling the electromagnetic response of complexstructures is not trivial. The basic modelling logic is depicted in Figure 6-1. Only with the vast improvement in computer power has it become possible to begin to address complex structures. It is therefore useful to put the history of GPR modelling into context.

Geophysical modelling, and GPR is no exception, has followed a traditional path. First, modelling of very simplestructures is carried out. Usually solutions of an analytical nature are considered (such as the earlier sections of thesenotes). Systems are simplified so that a closed form mathematical formula can be derived. In the GPR realm, 1Dmodels and simple analytical solutions were employed until the early 1990’s.

Figure: 6-1 This block diagram shows the general principles of the elements involved in forward modelling. Onehas to input the system characteristics combined with Maxwell’s equations and physical properties to derive an out-

put through a modeling algorithm based on these inputs.

8/20/2019 GPR Annan 2003


6-Modelling of GPR Responses GPR Principles, Procedures & Applications

128

As computer power permitted and software development allowed, movement to two-dimensional (2D) modellingoccurred. In many instances the modelling has been 2D because 2D modelling allows much insight with minimalcomputing cost. As a result, 2D modelling simplifies the real world into structures that vary in two directions and areinvariant in the third dimension (in geological terms called the strike direction of the structure). This has led to a vari-ety of numerical solutions such as finite difference and ray tracing solutions. Integral equation solutions although possible, have not seen much usage.

Finally, full simulation modelling incorporates all three dimensions. Complex three-dimensional (3D) structures canonly be modeled or simulated using numerical techniques. Techniques such as finite difference modelling (or finiteelementing modelling) or integral equation modelling are the norm but implementations have only appeared since themid 1990’s and then only with use of powerful computer systems. 3D modelling can be extremely complex, requiresaddressing the full vector solution of the fields, and is intensive both in memory and CPU requirements.

As PC’s become more powerful, more and more modelling of complex structures is occurring with less and less needfor finesse to speed up algorithms and reduce dimensions. GPR is still a long way from fast reliable solutions for complex 3D structures readily accessible to the average user community.

6.3 ONE-DIMENSIONAL (1D) MODELLING

6.3.1 R ADAR R ANGE EQUATION (RRE)

The radar range equation (RRE) is the simplest form of model that has been used in GPR for some time. The purposeof the radar range equation approach has been to assess the ability of a radar to probe into a material and detect anobject. It simplifies the 3D response using simplified parameterization. By estimating key system parameters, target parameters and material properties one can estimate depth of penetration and signal magnitude that can be expected.

The application of the RRE to GPR was first discussed by Annan & Davis (1977). This model has been used exten-sively for assessing the suitability of GPR and the possibility of it working in a particular situation. The major resultof RRE is recognizing how strongly attenuation impacts GPR. This solution is by no means a complete answer to any problem but it certainly gives insight into sensitivity and signal levels. Mathematically, RRE expresses the received

signal power which must be above the noise level to be detectable.

(6-1)

Figure 6-2 shows a block diagram of the radar range equation with various terms explained. The RRE breaks theresponse down into a series of steps starting from the energy created in the electronic source and through to thatwhich is detected at the receiving electronics. By following through these steps it is possible to estimate the magni-tude of the return signal which is likely to be observed. These solutions can be set up as computer programs or inspreadsheets to facilitate sensitivity analysis.

NOISE4

L-4

RXTXRXTXRX PL

e AgGGP >

πσξξ=

α

216

8/20/2019 GPR Annan 2003



129

Figure: 6-2 The radar range equation (RRE) is used to provide a simplistic estimate of response by determining power losses throughout the various steps of the measurement process.

6.3.2 ONE-DIMENSIONAL (1D) LAYERED EARTH

The 1D layered earth model is a carry over from the petroleum seismic industry. Generation of synthetic seismo-grams for layered earth environments provides a way to start to understand the effect of multiple paths, reverbera-tions, polarity changes and gradational interfaces.

1D GPR modelling is described by Annan & Chua (1992). The 1D model makes no claims to be quantitative aboutthe amplitude of the response that would be observed with a GPR. The model provides a sense of the timing of events and the relative amplitude of the signals. Issues such as the polarity of an event and the change in amplitude ata gradational boundary are GPR topics which have been addressed (Annan et al (1992), Redman (1994)). Figure 6-3demonstrates an example result of 1-D modelling. In this case a gradational boundary is being emulated to show the

decrease in amplitude as an interface becomes gradational with respect to the dimension of the excitation pulse.

8/20/2019 GPR Annan 2003



130

Figure: 6-3 One-dimensional GPR modeling involves defining the physical properties versus depth and then computing the response assuming no dimensionality to the source. This example on the right shows the physical model in

terms of relative permittivity and attenuation and the resulting synthetic GPR response on the left.

6.4 TWO-DIMENSIONAL (2D)MODELLING

6.4.1 R AY TRACING

For GPR, one of the ways to carry out 2-D modelling is to use the technique of ray tracing. In a ray tracing model one

assumes that the distances between objects is large and the pulse length duration is very short compared to the dis-tances involved. The signals can be treated using geometrical optics approximations. Such GPR modelling conceptsare again borrowed from the seismic field.

In a 2-D model, the responses can be broken down into electromagnetic field polarization where either the electricfield (TE) or magnetic field (TM) is parallel to the 2D structure. The rays representing the electromagnetic fields arethen traced through the model for the particular polarization type. The TE or TM separation reduces the vector complexity to a scalar problem. The rays are traced through the model and reflected and refracted at interfaces accordingto Snell’s law and the Fernel reflection coefficients. Amplitude attenuation along the ray paths follows the ohmic dis-sipation exponential decrease.

Ray tracing modeling is for GPR described by Cai & McMechan (1995). The ray-tracing concept for an undulatinginterface is depicted in Figure 6-4 and the GPR response is shown in Figure 6-5. The model response can be computed for either TE or TM field polarization.

8/20/2019 GPR Annan 2003



131

(a)

(b)

Figure: 6-4 Figure (a) shows the geometrical model for the ground structure. Figure (b) shows the vari-

ous ray paths from the surface down to the individual reflecting horizons.

Figure: 6-5 This shows a synthetic response computed for a model shown in Figure 6-4. In this case, the specularreflections from the various horizons appear at the appropriate delay times accounted for by the various velocities in

the section as well as the attenuation in the material.

8/20/2019 GPR Annan 2003



132

This type of model can incorporate the far-field directivity of the GPR antenna. Examples of insights into the behav-ior of geologic structures in complex situations using such a model are given by Nobes & Annan (2000).

6.4.2 FK 2D MODELLING

2D modelling also lends itself to finite difference solutions. One of the fast ways of computing the grid response toan arbitrary 2D structure is to transform space and time into the temporal frequency and a spatial frequency (or wave

number) domains. This transformation domain is called the FK domain, where F stands for frequency and K standsfor spatial wave number and implies double fourier transformation.

The modelling involves taking each interface or target in the model and turning it into a reflectivity coefficient at each point in the 2D space. As with the ray tracing approach, the vector to scalar reduction by TE and TM separation can be employed. The FK domain approach enables the propagation signals using simple analytical calculations andmanipulations. Once the propagation gets carried out in the FK domain, inverse fourier transform yields the space-time response.

While sounding complex, the calculations are fairly straightforward and the principles are extensively discussed inthe seismic literature. Application of 2D FK solutions to GPR problems are described by Zeng et al (1995).

An example model is shown in Figure 6-6 and the resultant FK solution is shown in Figure 6-7. In general the FK modelling has speed benefits for complex structures and includes diffracted energy whereas ray tracing solutions dealonly with specular reflection energy but do not address diffraction response. Ray tracing generally allows for physi-

cally larger problems because only the boundaries need to be considered.

Figure: 6-6 When carrying out a two dimensional FK model, one places reflectivity at every point on a regularmesh. The model shown here is an example of reflectivity versus position.

Figure: 6-7 This section shows the result of FK 2-D modeling output for the reflectivity model shown in Figure 6-6 .

8/20/2019 GPR Annan 2003



133

6.4.3 FULL FINITE DIFFERENCE (FD) SOLUTIONS

The first two 2D solutions discussed here are approximate solutions. Full 2D numerical simulation for either time(FDTD) or frequency (FDFD) domain is practical. TE and TM separation again allow reduction of the vector prob-lem to a scalar calculation. Very little of 2D FD has occurred in the GPR field as the advent of 3-D modellingoccurred at the same time as 2D modelling. Unlike the seismic and other geophysical modelling worlds where computer technology just could not handle 3D modelling for many years, demand for 3D GPR modelling began at the

same time computer power allowed it to occur (see Giannopoulos (1997)).

The importance of 2D FD modelling is that it can be fit into a moderate size computer and is a scalar calculationrather than a vector solution. The vector solution can be built up by adding the two types of responses which areimplicitly vector fields but have no mixing.

6.4.4 2½ D SOLUTIONS

In some situations, including the localized 3D nature of the source of the electromagnetic fields is desirable for ground models which are two-dimensional. 2½ D modelling exploits the 2D solution by splitting the excitation in TEand TM components, solving the 2D problem and combining the results at the end. As with simpler 2D, little GPR work in this area has occurred. 2½ D is more common in traditional EM (Hohmann 1987). A good GPR example isthe work of Lampe and Holliger (2000).

6.5 THREE-DIMENSIONAL (3D) MODELLING

6.5.1 3D R AY TRACING

3D ray tracing solutions have a long history in seismic (Kline and Kay (1965), Cerveny and Ravindra (1971)) andelectromagnetics (Luneburg (1964), Borne and Wolf (1980)). In these cases, the signals propagate large distancesthrough the material compared to the wavelength and the ray optics approximation is quite suitable in manyinstances.

For GPR, 3D ray tracing solutions have seen limited use and development. There are two reasons for this. First,GPR signals are strongly attenuated limiting the conditions where the geometric optics assumptions are fully applicable. Secondly, by the time the demand for 3D modelling started to occur, computer power had reached the levelwhere it was possible to consider full numerical solutions and hence the use of approximate solutions, such as raytracing which becomes complex in 3D, were perceived to have little benefit.

6.5.2 3D FINITE DIFFERENCE MODELLING

The majority of 3D GPR modelling uses numerical finite difference style techniques. Depending on the particular application, the solution may be set up as finite element or finite difference but the general implication is that one issolving the full vector field at a mesh of points in 3D space.

The initial 3D modelling for GPR was reported by Roberts & Daniels (1996). 3D modelling for EM induction wasalso becoming available Alumbaugh and Newman (1994). This modelling required use of a Cray super computer togenerate results. Similar modelling results are reported by Bergmann et al (1996), Wang & Trip (1996), Carcione(1996) and Holliger & Bergmann (2000).

It is beyond the scope of this brief discussion to get into the details of numerical simulation. Referring to the above papers is the best way to learn about 3D GPR modelling. There are a number of different techniques which are used

to approach these problems and considerable skill and modelling experience is required to confidently enter this areaof GPR at present. One must solve the vector field equation for the field components with full variation of the mate-rial properties.

One can solve for the fields for particular sinusoidal excitation frequency. This approach was the most commonmethod to start with as commonly returned to as FDFD (finite difference frequency domain) modelling.

Using a frequency domain solution, the time domain solution is obtained by modelling at a number of different fre-quencies and then fourier transforming the frequency response into the time domain to obtain the transient response.

8/20/2019 GPR Annan 2003



134

More common now is the FDTD (finite difference time domain) solution. The approach here is to by-pass the fre-quency domain and solve the full transverse vector wave equation in the time domain. These results require both spa-tial and temporal finite differencing or numerical discretization. To date, however, the complexity of the responsesand the length of time required to compute them has limited widespread use. As computer power increases and computer codes become more efficient, more and more utilization of the numerical solutions is occurring.

The key reason why full 3D numerical solutions are needed is that they provide the basis for inverting or extracting

quantitative information about the subsurface in an automatic fashion.

Figure: 6-8 This grid pattern shows the concept of the staggered grid for electromagnetic modeling. The electric

and magnetic fields are computed independently at grid points which are off set from one another. The magnetic fields at points around an electric field point are used to update the electric field. Similarly the electric field points

around an individual magnetic observation point are used to update a magnetic field.

The general approach to 3D modelling is to establish a grid. If a single vector field (i.e. electric or magnetic) is usedthen a second order spatial difference equation is required. More commonly, a staggered grid, where both the electricand magnetic field are computed alternately in time using first order derivatives rather than second derivatives as proposed by Yee (1965), has become the standard in electromagnetic modelling.

The example grid shown in Figure 6-8 shows a staggered grid. The electric field is computed on one set of meshnodes while the magnetic field is computed at another set of mesh nodes which are offset.

Once these solutions are computed they can be displayed as fields versus time such as the 3D wavefront shown in

Figure 6-9. These examples were drawn from the paper by Giannopoulos (1997).

8/20/2019 GPR Annan 2003



135

Figure: 6-9 An example of using a finite difference time domain (FDTD) computational algorithm based on a staggered grid to compute the response of a reinforcing bar buried in concrete.

Other topics which are addressed in FDTD models are the inclusion of dispersive electric properties (Bergmann et al(1999), Young and Nelson (2001)). In all numerical models, the edges or boundaries of the model space can createartificial events. Delving into the modelling field requires learning about devising absorbing boundary conditions(Holtzman and Kastner (2001), Oguz, Levent (2001)).

6.5.3 INTEGRAL EQUATION – EQUIVALENT SOURCE SCATTERING

The equivalent source method is an elegant and physically intuitive means of formulating EM responses (Hohmann(1987), Annan (1974)). Instead of numerically solving for all the fields, the approach was the known response andformulates the response for localized targets which can greatly reduce the size of the numerical calculation.

Figure: 6-10 The conceptual idea of an embedded localized inhomogeneity (a) being transformed into an equiva-lent signal source (b) in a known background.

8/20/2019 GPR Annan 2003



136

Details of this methodology and beyond this presentation. The concept is explained in scalar form here. Referring toFigure 6-10, a localized target with differing properties is present in a media of a known response which can be char-acterized by the green’s function

(6-2)

If a source is present and described in space and time as s(r,t), the fields created, f(x,t), are mathematically expressedas the convolution

(6-3)

By describing the target as a difference in physical properties p = (p-pB) incorporated in the background responsegreens function, the excitation field f(r,t) will cause an apparent source signal

. (6-4)

This new equivalent source generates the response

(6-5)

with the source signal satisfying an integral equation of the form

(6-6)

when the material contrast p is small, the approximation is used.

. (6-7)

This is called the Born approximation and enables fast calculation of responses.

Because this solution has a form which can be computed in direct manner where approximated, the inversion proce-dures currently common in GPR (and most EM) use this modelling approach as the basis for structuring inversionschemes. For this reason, more detail is provided here than on other modelling.

6.6 INVERSION

Inversion is the mathematical technique used to describe the extraction of physically quantifiable parameters out of aset of observed data. In concept, a measurement system obtains data at a number of spatial points which containencompass a number of different frequencies or different time delays (i.e. band limited transfer function or impulseresponse). This information is indicative of targets or structures buried in the ground. The goal is to infer the spatialdistribution of the material properties which give rise to the observed response.

)tt,,r (r,g =

∫∫∫ ′′′′′′= tdr )dtr )s(tt.,r g(r,t)f(r, 3

∆

))t,r (f )t,r (f ( p)t,r (f p)t,r ( eTOTAL +∆=∆=es

∫∫∫ ′′′′′′= tdr )dt,r ()stt,,r g(r,t)(r,f 3ee

∫∫∫ ′′′′′′∆+∆= tdr )dt,r ()stt,,r g(r,pt)pf(r,t)(r,s 3ee

∆

)t,r ( pf )t,r ( ∆≈es

8/20/2019 GPR Annan 2003



137

Figure: 6-11 Inversion of GPR data requires forming the opposite steps as discussed for Ford modelling. Given themodel, the algorithm and the observation of ground response, one reverses the modelling steps to extract the physical

property description.

The forward modelling described provides the tools whereby the responses can be predicted if a known distri butionof physical properties is given. Inversion is the reverse of this process. The concepts are shown figuratively in Figure6-11. Given the observed fields these observations must be transformed back into a spatial distribution of physical properties which are hopefully unique and dependable. The subject of inversion is well known in the geophysicalcommunity and excellent discussions of the topics can be found in Tarantota (1987) and Oldenburg et al (1993).

The first successful extraction of physical properties from GPR is described by van der Kruk (2001). In this work thevector fields at the surface as a function of time are processed to extract changes in permittivity and conductivity inthe subsurface. The solution is formulated using the Born approximation integral equation solution. This work pro-vides an excellent basis for understanding the future evolution of inversions of GPR responses into quantifiable mate-rial properties.

To date, the approach in GPR has been to estimate or image reflectivity versus spatial location in the subsurface. Anintroductory discussion is provided by Annan (1997). Techniques of migration and synthetic aperture are applied asopposed to full inversion. The reflectivity imaging concepts are well developed in the seismic field. An insightfuldiscussion on this is given by Bleistein and Gray (2001). Discussion of the GPR aspects and the handling of someaspects of the vector character are discussed by Green et al (2002), Moran (2000).

8/20/2019 GPR Annan 2003



138

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 7-Survey Design

139

7 SURVEY DESIGN

Proper design of GPR surveys is critical to success. Setting expectations and optimizing data acquisition to meetexpectations requires planning. This chapter builds on the paper Annan & Cosway (1992) on the subject of surveydesign.

7.1 EVALUATING GPR SUITABILITY

Prediction of whether GPR will "work" for the problem at hand is not clear cut. In general it is easier to rule out situ-ations where radar is totally unsuitable than to state with confidence that radar will be successful. Again, this is not aunique feature of the GPR method but is a fact of life with all geophysical methods. GPR tends to have more mystery because people have not normally had as much experience with it as with some other methods.

There are some basic tools which assist the GPR user in the decision making process. The two most important are theradar range equation, and numerical simulation techniques. Some examples are described by Annan and Chua(1988).

The radar range equation (RRE for short) does a basic allocation of available power against all the loss mechanismsto yield a "yes/no" answer on whether a target will return sufficient power to be detectable. The RRE has to simplifythe problem at hand; therefore, the results are good guides, not absolute predictors of success or failure. The basicsteps of the radar range equation are depicted in Figure 7-1. Example results of an automated program to carry out

these calculations are shown in Figure 7-2. Nomograms for specific systems and targets can also be generated suchas shown in Figure 7-2.

Figure: 7-1 Block Diagram of radar range equation.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003


7-Survey Design GPR Principles, Procedures & Applications

140

System Performance Factors

System Q (dB) 133.98 Tx voltage (V) 1000.00Tx to antenna effcy (dB) -20-00 Rx noise (uV) 200.00Transmitter antenna gain (dB) 3.00Antenna to receiver effcy (dB) -20.00Receiver antenna gain (dB) -1.45 _____

Net system Q (dB) 98.53

Propogation Factors

Spreading losses (dB) -74.02Attenuation Losses (dB) -4.00

Target Factors POINTS ROUGH SPEC SPEC THIN Target

backscatter fain (dB) -5.37 8.60 19.85 -38.72

Net Performance 15.14 29.11 40.36 -18.21

Target amplitude (uV) 1143 5709 20853 25

Stacks 1 1 1 66

Windows (ns 667 RADAR RANGE PAPER COEFFTS

Model Parameters POINT ROUGH SPEC PEC THIN

Centre Freq. (MHz) 100.00 B1-8.54 1.56 0.62 -6.47Target range (m) 20.00 B20.00 1.00 2.00 2.00Attenuation (dB/m) 0.10 B34.00 -1.00 0.00 0.00Target diameter (m) 0.50

Target Reflection (dB) -11.14Host K 25.00

*Thin layer reflection (dB) -69.71Target K 8.00

Layer Thickness (m) 0.00 *Thin layer beta (dB) -58.57

Conductivity Attenuation *Diameter/wavelength (dB) -1.58

5.00 mS/m 1.69 dB/M Wavelength (m) 0.60

* If these values exceed -10 dB then, the assumption and approximations used to derive these formulas are invalid.

Figure: 7-2 Example of a radar range calculation.

Numerical simulation techniques (NST for short) are now becoming well developed for GPR. Simple programs for2D earth structures are commercially available and are instructive to use. More complex 2- and 3-dimensional mod-elling programs are not available for general use but are available at research institutions. The basic concepts for plane (flat) layered earth modelling are shown in Figure 7-3 accompanied by an example synthetic generated by thecommercial Sensors & Software Inc. synthetic radargram program. Simple 2D earth responses using ray tracing andfinite difference approaches are shown in Figure 7-4. (McMechan) (1981)

8/20/2019 GPR Annan 2003



141

Figure: 7-3 Radar range equation nomogram example.]

Figure: 7-4 Illustration of a synthetic radargram to predict a GPR response.

8/20/2019 GPR Annan 2003



142

Figure: 7-5 Modeling such as this 2 dimensional ray trace stimulation helps users utility of GPR .

Answering the question "Will GPR work???" is neither easy nor exact. Addressing the following three questions willcertainly help in anticipating the answer.

Question 1: Is the target within the detection range of the radar irrespective of anyunusual target characteristics?

Figure: 7-6 What is the target depth?

8/20/2019 GPR Annan 2003



143

Discussion of Question 1

The way to answer this question is to calculate or measure the host attenuation coefficient. Using the radar rangeequation and the system performance factor (example inFigure 7-2), compute the maximum range that a reflector ofthe anticipated target type can be detected. If the target is at a depth greater than this range, radar will not be effec-tive. A conservative rule-of-thumb is to state that radar will be ineffective if the actual target depth is greater than50% of the maximum range.

Commercial radar systems can typically afford to have a maximum of 60 dB attenuation associated with conductionlosses. A rough guide to penetration depth is

(7-1)

where α is attenuation in dB/m and å is conductivity in mS/m. These equations are not universal but are applicablewhen attenuation is moderate to high α> 0.1 dB/m or σ > 1 mS/m which is typical of most geologic settings.

Question 2: Will the target generate a response detectable above the backgroundclutter? In other words, does the target have sufficient contrast in electrical prop-erties and is it physically large enough to reflect or scatter a detectable amount of

energy?

Figure: 7-7 What is the target geometry?

Figure: 7-8 What are the target electrical properties?

σ

35= D

8/20/2019 GPR Annan 2003



144

Figure: 7-9 What is the host material?


Power reflectivity is estimated using the expression

(7-2)

Two conservative rules-of-thumb for predicting success are as follows. First, the electrical properties of the targetshould be such that the power reflectivity be at least 0.01. (Note that a metal target is equivalent to K Target→∞.) The

size of the target also is a factor in the amount of energy scattered. While specific shape is also important, the sizeeffect dominates and can be best seen from the scattering cross section of a sphere versus wavelength shown inFigure7-10. For targets small with respect to the wavelength, scattering cross section depends on wavelength. For targetslarger than the wavelength, the cross section stabilizes to a constant.

Figure: 7-10 Variation of the scattering cross section of a spherical target as a function of dimension normalized

against wavelength (diagram from Skolnik(1970)). While specific to a sphere, similar behaviour is displayed by any

object of finite dimension. The response increases until the object is on the same order of size as the wavelength. For

GPR to penetrate through a heterogeneous material, it highly advantageous that the GPR wavelength be large com-

pared to the scale of heterogeneity.

2

TargetHost

TargetHost

+

−=

K K

K K P r

8/20/2019 GPR Annan 2003



145

Question 3: Can you confirm that other obvious conditions will not preclude useof GPR?

Figure: 7-11 What is the survey environment like?


One example would be a radio transmitter located at the site. Another example would be a tunnel lined with metalmesh to prevent loose rock from falling. In the first case external signals may saturate the sensitive receiver electron-ics. In the later, all the radar signal would be reflected at the tunnel wall and none would penetrate into the tunnelwall.

If the answers to all the above questions are "yes", there is a good chance GPR will work. The above conditions are posed in a conservative manner and intended to err on the pessimistic side. More detailed analyses can employ RREand NST techniques. In general it is almost impossible to obtain reliable estimates of all of the parameters involvedin RRE and NST; these procedures are most effectively used as part of a sensitivity analysis. The conclusions drawnwill be fuzzy but informed.

As with all predictions nothing beats a real field trial. If practical, a field evaluation stage should be an integral component in survey design optimization. Unfortunately, financial constraints usually are a real and limiting factor.

7.2 R EFLECTION SURVEY DESIGN

The most common mode of GPR surveying is common-offset, single-fold reflection profiling as depicted in Figure 7-12. In such a reflection survey, a system with a fixed antenna geometry is transported along a survey line to mapreflections versus position.

Figure: 7-12 Schematic illustration of common-offset single-fold profiling.

8/20/2019 GPR Annan 2003



146

There are seven parameters to define for a common-offset, single-fold GPR reflection survey. These are the fre-quency, the time window, the time sampling interval, the station spacing, the antenna spacing, the line location andspacing, and the antenna orientation.

7.2.1 SELECTING OPERATING FREQUENCY

Selection of the optimal operating frequency for a radar survey is not simple. There is a trade off between spatial res-olution, depth of penetration and system portability. As a rule, it is better to trade off resolution for penetration.There is no use in having great resolution if the target cannot be detected!! The following is a brief summary of moreextensive discussions of this subject given by Annan & Cosway (1994).

There are three main issues which control frequency selection, namely,

i. Spatial resolution desired,

ii. clutter limitations, and

iii. exploration depth.

Each of these issues yields a constraint on frequency. A brief description of each topic is presented and the frequencyconstraint given without detailed derivation.

Resolution of two events requires that the radar pulse envelope time duration be shorter than twice the separationdelay time between two features to be resolved. Assuming a centre frequency to bandwidth ratio of 1, the constrainton the centre frequency, f c, takes the form

(7-3)

where ∆Z is the spatial separation to be resolved in metres and K is the dielectric constant or relative permittivity. Inother words, spatial resolution places a lower bound on the centre frequency. (and required bandwidth).

Clutter in GPR systems refers to the radar signals returned from material heterogeneity in soils and rock. The radar

response of small scale features (i.e., fine scale bedding, cracks and joints, laminations), increases rapidly as radarfrequency increases as evident in Figure 7-10. The data example in Figure 7-13 clearly show how clutter increaseswith increasing frequency). If the radar frequency becomes too high, one can often reach the point where one "can'tsee the forest for the trees"!! In order to "see" to depth into the ground, the amount of energy scattered by cluttershould be minimized. To achieve this, the signal wavelength should be much longer (we use a factor of 10) than thetypical heterogeneity or clutter dimension, ∆L, in the host environment. The clutter centre frequency constraint takesthe form

(7-4)

Note that this constraint also implies that the target sought must be considerably larger than the clutter dimension. Ifthis is not true, then identifying the target response becomes the problem of looking for the "needle in the haystack".

z75

ΜΗΚ ∆Ζ

> R

c f

z30

ΜΗ∆

> K L

f C

c

8/20/2019 GPR Annan 2003



147

Figure: 7-13 The above data sets clearly illustrate the clutter frequency concept. These data were acquired over two

tunnels in an area of gneissic bedrock. The rock texture had a spatial scale of 30 cm. At 100 MHz the clutter is clearly

visible. At 50 MHz much of the clutter from the rock texture is suppressed.

The third frequency, referred to as the exploration depth frequency, requires that the target cross section occupy amajor fraction of the radar beam in order that sufficient energy be returned for detection. Furthermore, the targetdimension should be as close in size as possible to a Fresnel zone in order that the returned signal arrive coherently.Derivation of a simple guide to encompass these trade offs without resorting to a full radar range analysis is not simple but a basic constraint on the centre frequency is that

(7-5)

where β is the ratio of radar beam footprint to target size ratio and D is depth in metres. An assumption of β = 4 isreasonable for GPR applications and one obtains

(7-6)

as a centre frequency constraint.

As a general rule, the three frequencies should be computed and, if the survey problem has been properly posed, oneshould find that

(7-7)

If the resolution frequency is greater that the depth or the clutter frequency, the desired spatial resolution is incompat-ible with the clutter dimension or depth of exploration.

D K f D

c1v −< β

MHz11200

D

K f

D

c

−<

),min( C

c

D

cc

R

c f f f f <<

8/20/2019 GPR Annan 2003



148

A quick guide to frequency selection is to use Table 7-1 which is based on the assumption that the spatial resolutionrequired is about 25% of the target depth.

Table: 7-1 Antenna centre frequency as a function of exploration depth.

The above are values based on practical experience. Since every problem requires careful thought, the above valuesshould only be used as a quick guide and not a replacement for thoughtful survey planning.

7.2.2 ESTIMATING THE TIME WINDOW

The way to estimate the time window required is to use the expression

(7-8)

where the maximum depth and minimum velocity likely to be encountered in the survey area are used. The aboveexpression increases the estimated time by 30% to allow for uncertainties in velocity and depth variations.

If no information is available on the electrical properties. Table 7-2 provides a quick guide.

Table: 7-2 Time Windows

Depth (m)Center

Frequency (MHz)

0.5 1000

1.0 500

2.0 200

7.0 100

10.0 50

30.0 25

50.0 10

Depth (m) Rock Wet Soil Dry Soil

0.5 12 24 10

1 25 50 20

2 50 100 40

5 120 250 100

10 250 500 200

20 500 1000 400

50 1250 2500 1000

100 2500 5000 2000

Velocity

Depth21.3W ×

=

8/20/2019 GPR Annan 2003



149

7.2.3 SELECTING TEMPORAL SAMPLING INTERVAL

One of the parameters utilized in designing radar data acquisition is the time interval between points on a recordedwaveform. The sampling interval is controlled by the Nyquist sampling concept and should be at most half the periodof the highest frequency signal in the record. For most ground penetrating radar antenna systems, the bandwidth tocentre frequency ratio is typically about one. What this means is that the pulse radiated contains energy from 0.5times the centre frequency to 1.5 times the centre frequency. As a result the maximum frequency is around 1.5 timesthe nominal centre frequency of the antenna being utilized.

Based on the assumption that the maximum frequency is 1.5 times the nominal antenna centre frequency, the datashould be sampled at a rate twice this frequency. For good survey design it is better that one uses a safety margin oftwo. As a result the sampling rate should be approximately six times the centre frequency of the antenna being uti-lized. Based on this analysis Table 7-3 summarizes suitable sampling intervals versus operating frequency when oneassumes a centre frequency to bandwidth ratio of 2

Table: 7-3

The function relationship is

(7-9)

where f c is the centre frequency in MHz and t is time in ns.

In some instances it may be possible to increase the sampling interval slightly beyond what is quoted but this shouldonly be done when data volume and speed of acquisition are at a premium over integrity of the data. In any event thesampling interval should never be more than 2 times that quoted here.

7.2.4 SELECTING STATION SPACING (SPATIAL SAMPLING INTERVAL)

The selection of spacing between discrete radar measurements (see Figure 7-7) is closely linked to the centre operat-ing frequency of the antennas and to the dielectric properties of the subsurface materials involved. In order to assurethe ground response is not spatially aliased, the Nyquist sampling intervals should not be exceeded. The Nyquistsampling interval is one quarter of the wavelength in the host material and expressed as

(7-10)

Antenna Center

Frequency (MHz)

Maximum Sampling

Interval (ns)

10 16.70

20 8.30

50 3.30

100 1.67

200 0.83

500 0.33

1000 0.17

c f t

6

1000=

)m(in75

4 K f K f

c x ==∆

8/20/2019 GPR Annan 2003



150

where f is the antenna centre frequency (in MHz) and K is the relative permittivity of the host. If the station spacingis greater than the Nyquist sampling interval, the data will not adequately define steeply dipping reflectors or diffrac-tion tails. In areas of flat lying reflectors, this criteria can be compromised.

The spatial interval of measurement is clearly illustrated by the example sections shown in Figure 7-14 and Figure 7-15. The 50 MHz data in Figure 7-14 were collected with a station spacing of 3m which is considerably larger than thecomputed Nyquist interval of 0.5 to 0.77. The data in Figure 7-9 clearly define the strong relatively flat lying reflec-

tors. The steeply dipping events are aliased and appear as 'hash' on the section. The same section sampled at a 0.5mstation interval shown inFigure 7-15 clearly defines the steeply dipping features.

Figure: 7-14 GPR reflection section from a deltatic environment obtained using 50 MHz antennas and a 3m station

spacing.

Figure: 7-15 Same section as Figure 7-9 but station interval reduced to Nyquist sampling interval of 0.5 m.

There are practical trade-offs to be made in selection of station interval. From a practical viewpoint, data volume andsurvey time are reduced by increasing the station interval. From a data interpretation standpoint, adhering to the Nyquist sampling interval is very important. There is also nothing to be gained by spatial oversampling. The sampling interval is extremely important as this example indicates and should be carefully weighed in the survey design process.

The visual aspect of data to the human eye is also an important factor. Data which are adequately simple may appearslightly blocky. Suitable interpolation in the data display program can remedy this effect.

8/20/2019 GPR Annan 2003



151

7.2.5 SELECTING ANTENNA SEPARATION

Most GPR systems employ separate antennas for transmitting and receiving (commonly referred to as bistatic opera-tion) although the antennas may be housed in a single module with no means of varying the antenna separation. Theability to vary the antenna spacing can be a powerful aid in optimizing the system for specific types of target detec-tion.

To maximize target coupling, antennas should be spaced such that the refraction focussing peak in the transmitter andreceiver antenna patterns point to the common depth to be investigated. Since the antenna pattern peaks at the criticalangle of the air-earth interface as illustrated in Figure 7-16, (Annan, 1973, Annan et al, 1975 Smith, 1984). An esti-mate of optimum antenna separation is given by the expression

(7-11)

Increasing the antenna separation also increases the reflectivity of flat lying planar targets which can sometimes beadvantageous.

Figure: 7-16 Variation in antenna pattern as relative permittivity of the ground changes. (Upper medium is air).

A very practical reason for increasing antenna separation is receiver dynamic range. The direct transmitter toreceiver signal can be very large if the antenna separation is small. While electronic circuits may handle the overloadsafely, there is distortion and recovery time which make detection of shallow events impossible (transmit pulse blank-ing is the term commonly used). As minimum separation of a half-wavelength at the centre frequency is recom-

mended.

A factor which should be considered when working in lossy ground is the effect of path length increase created byseparating antennas. The path length is

(7-12)

In lossy ground, return signals are attenuated by a factor

1

Depth2

−=

K S

2

122 )Depth4S( += L

8/20/2019 GPR Annan 2003



152

(7-13)

One should not make S so large that L becomes a great deal larger than 2 x Depth in high loss conditions.

If little is known about the survey area, a safe rule-of-thumb is set S equal to 20% of the target depth. In practice, asmall antenna spacing is often used because operational logistics usually demand simplicity of operation. Depth res-

olution of targets decreases as antenna separation increases although this factor is small until S approaches the targetdepth.

7.2.6 SURVEY GRID AND COORDINATE SYSTEM

An important aspect of survey design is establishment of a survey grid and coordinate system. The use of a standard-ized coordinate system for position recording is very important; the best data in the world are useless if no one knowswhere they came from.

Generally, survey lines are established which run perpendicular to the trend of the features under investigation inorder to reduce the number of survey lines. Line spacing is dictated by the degree of target variation in the trenddirection. If isolated small targets are sought, the line spacing should be less than the radar footprint illustrated inFigure 7-17.

Figure: 7-17 Simplified GPR footprint concept where shaded zone depicts area illuminated at depth. (Annan and

Cosway, 1992).

The selection of survey line location and orientation should be made such as to maximize target detection. All surveylines should be oriented perpendicular to the strike of the target if the target has a preferred strike direction. Inattempting to cover an area to map a feature such as bedrock depth, the survey lines should be oriented perpendicularto the bedrock relief and line spacing should be selected to adequately sample along-strike variations without alias-ing. In situations where strike is known and the structure 2-dimensional, a very large spacing between lines can beemployed. If there is no two dimensionality to the structure, then line spacing must be the same as the station spacingto assure that the ground response is not aliased. Needless to say, when ∆x is a fraction of a meter, a tremendousamount of data have to be collected to fully define a 3-dimensional structure.

Le α −

8/20/2019 GPR Annan 2003



153

7.2.7 SELECTING ANTENNA ORIENTATION

The last factor and seldom discussed factor to be considered is the antenna orientation. In general, the antennas usedfor GPR are dipolar and radiate with a preferred polarity. The antennas are normally oriented so that the electric fieldis polarized parallel to the long axis or strike direction of the target. There is no optimal orientation for an equi-dimensional target. In some instances, it may be advisable to collect two data sets with orthogonal antenna orienta-tions in order to extract target information based on coupling angle. If the antenna system is one which attempts touse a circularly polarized signal, the antenna orientation becomes irrelevant. Since most commercial systems employ

polarized antennas, orientation can be important. The various arrangements of antenna deployment are illustrated inFigure 7-18. The most commonly used is the one designated 'PR-BD'.

Figure: 7-18 Illustration of the various modes for antenna deployment. (E field assumed aligned along the antenna

axis

Antenna orientation affects the subsurface footprint size. As the simplified beam pattern (see Figure 7-17) indicates,the simple dipole antenna has a broader footprint in the "broad side" direction than in the "endfire direction". Thefield example shown inFigure 7-19 graphically illustrates the difference in pattern widths. For this reason, the 'PR-

BD' survey configuration tends to give a radar section which is more of a 2D slice through the subsurface and is notsubject to a large amount of "side swipe" clutter from objects off line.

Figure: 7-19 CMP sounding data acquired with 2 modes of antenna polarization.

8/20/2019 GPR Annan 2003



154

For surveys where interpretation of data in a quantitative manner is important, this result is beneficial. In search anddetect surveys, a broader search swath obtained by the 'PL-PD' and 'PL-EF' may be more desirable.

7.3 CMP/WARR VELOCITY SOUNDING DESIGN

The CMP (common mid-point) and WARR (wide angle reflection and refraction) sounding modes of operation are

the electromagnetic equivalent to seismic refraction and wide angle reflection. CMP/WARR soundings are used toobtain an estimate of the radar signal velocity versus depth in the ground by varying the antenna spacing at a fixedlocation and measuring the change of the two-way travel time to the reflections as illustrated in Figure 7-20 and Fig-ure 7-21.

Figure: 7-20 Procedure for conducting a CMP velocity sounding.

Figure: 7-21 Illustration of CMP sounding ray paths and idealized event arrival-time versus antenna separation

and display.

8/20/2019 GPR Annan 2003



155

In the CMP sounding, both the antennas are moved apart about a fixed location. In a WARR sounding, one antenna isheld fixed while the other is moved away. In early GPR systems utilizing metallic cables, the WARR approach ofsounding was preferred because cable handling was a major concern in obtaining data free of system artifacts. Withmodern systems such as those employing fibre optics cables, the CMP approach is the standard mode of operationsince the reflected signal is more likely to come from a fixed spatial location rather than moving about on the surfaceof a reflector as occurs with a WARR sounding.

Optimal CMP/WARR soundings are obtained when the electric fields of the antennas are parallel and the antennas aremoved apart along a line which is perpendicular to the electric field polarization (the 'PR-BD' configuration in Figure7-18). This configuration gives the widest angular coverage of a subsurface reflector. In addition, close coupling ofthe antennas to the ground should be maintained in order to maximize reflection energy detectable at angles beyondthe critical angle of the air-earth interface.

The procedure for a CMP/WARR sounding is simple. A reflector is normally identified from a reflection section. A point on the ground surface is selected which is over the reflector. Antennas are then positioned over the target pointwith minimal separation. The initial spacing is usually nx , the Nyquist station interval selected for reflection profil-ing. Data are then acquired at antenna separations which increase as integer multiples of nx. If the CMP mode isused, both antennas are moved out from the centre point in steps of nx/2. If WARR mode is used, one antenna ismoved out in step intervals of nx . The maximum separation in a CMP/WARR sounding is usually 1 to 2 times thereflector depth. If the ground attenuation is high, the signals may die out before the maximum separation is reached.

The reflection arrival times should have a hyperbolic dependence (to first order) on antenna separation. Example

data sets are shown inFigure 7-19 for both the PR-BD and PL-EF antenna configurations. Analysis of the move outhyperbola of time versus separation permits estimation of propagation velocity and target depth. The basic interpre-tation procedure is "T2 - X2 " analysis commonly used in early seismic reflection interpretations. In simple terms, a plot of travel time squared versus antenna separation squared yields a straight line relationship whose slope gives avelocity estimate and whose time intercept yields a depth estimate (see Figure 7-22). Computerized schemes of vary-ing degrees of complexity are now commonly used to do this type of analysis such as the velocity stack shown in Fig-ure 7-23.

Figure: 7-22 T2-X2 analysis of CMP/WARR sounding.

8/20/2019 GPR Annan 2003



156

Figure: 7-23 Example of moveout velocity stacking of a PR-BD data set.

CMP/WARR soundings provide a measure of signal attenuation in the ground although to date the estimation is morequalitative than quantitative. In addition, at sites with a large amount of surface clutter, CMP/WARR soundings canaid in separating above and below ground reflections.

While not discussed in detail here, one can also conduct multifold GPR reflection surveys. By measuring reflectiondata at a multiplicity of offsets along a line, CMP data is available at each station and a full velocity section can bederived. These type of data are very important when large velocity variations occur and migration processing is to beused to reconstruct more representative depth sections from GPR data. Papers by Fisher et al (1992 a & b) andGrieves (1998) illustrate these concepts.

7.4 3D SURVEY DESIGN

Refer to the Annan et al. paper, Maximizing 3D GPR Image Resolution: A Simple Approach, 1997.

7.5 BOREHOLE SURVEY DESIGN

Refer to the Annan et al. paper, Cross Hole GPR for Engineering and Environmental Applications, 1997.

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 8-Data Processing

157

8 DATA PROCESSING

Ground penetrating radar (GPR) is now used extensively for a variety of applications in many differing fields. Theubiquitous access to inexpensive computer facilities means that more and more computer processing of GPR dataoccurs. The objective here is to provide a brief overview of GPR data processing. The very broad nature of the topicmakes it impossible to provide a complete discussion. The focus, will be to indicate the steps which should be fol-lowed in data processing and to stress the need for the processor (the person not the computer) to remember that there

must be a cost benefit at the end of the day.

Processors (the people not the computers) can exploit many of the developments of seismic data analysis which haveevolved to a very high level. Numerous commercial software packages are available which allow almost any imagin-able data manipulation. For anyone contemplating use of seismic processing on GPR data, the excellent text by Yil-maz (1987) should be considered essential reading. While all seismic processing can not be applied to GPR data, thevast majority can be used directly as evidenced by Fisher et al (1992a), Maijala (1992), and Rees & Glover (1992).

GPR data are most often treated as a scalar quantity while the electromagnetic fields which are the basis of themethod are vector quantities. Extensive use of GPR where the vector nature of the field has been exploited are few but growing. For processors with a seismic background, GPR data are more analogous to shear wave than compres-sional wave data.

The processing flow for GPR data is depicted in Figure 8-1. The initial stage of operations is the acquisition of dataand this is normally accompanied by real-time display. In many applications the real-time display is used for on-site

interpretation and may indeed be the end point for the radar survey. Frequently data are now recorded and are avail-able for post-acquisition processing and re-display. The highlighted zone in Figure 8-1 is the topic of this paper. Theareas of data processing have been grouped under the headings: data editing, basic processing, advanced processing,and visualization/interpretation processing. Processing is usually an iterative activity. A data set will flow throughthe processing loop several times with the data changes visually monitored by the processor (i.e. the person). Straightthrough batch processing may be applied on large data sets after iterative testing on selected data samples.

Figure: 8-1 A general overview of GPR data flow.

The data processing subdivisions indicated are meant to reflect that the data manipulation becomes more and more processor (i.e. the person) dependent and hence more subjective because a particular end result is sought. The fol-lowing discussion is biased toward single fold, common offset, reflection profiling data processing owing to the factthat the majority of GPR data is of this type.

The issue of data display, hard copy and visualization while somewhat beyond the scope of data processing is aninherent factor in data processing. Raw information, which is normally an amplitude versus time signal with severaldecades of dynamic range, must be manipulated for the display devices which have about one decade of dynamicrange. Since so many formats such as color mapping, grey scale mapping and line graphs are available, it is difficultto totally divorce processing from presentation considerations.

8/20/2019 GPR Annan 2003


8-Data Processing GPR Principles, Procedures & Applications

158

8.1 DATA EDITING

Once data are recorded, the first step in processing is data editing. Field acquisition is seldom so routine that noerrors, omissions or data redundancy occur. Data editing encompasses issues such as data re-organization, data filemerging, data header or background information updates, repositioning and inclusion of elevation information withthe data.

While this may sound like a trivial exercise, it is often the most time consuming one for a production survey. Largevolumes of information have to be systematically maintained so that further processing can be done without having toconstantly account for idiosyncrasies of acquisition. In many instances background information on acquisition isused in the processing stages. For instance, the centre frequency of the acquisition system antennas is a very impor-tant factor in processing. The temporal and spatial sampling interval are also important in most advanced processingsteps. Data recorded on equal spatial and temporal time intervals are virtually mandatory for most of the advanced processing methods. As a result, data editing is essential before further processing in many situations.

Figure 8-2 illustrates some of the many activities involved with data editing. The upper portion of the figure shows atypical file header which contains initial survey information including the number of traces in the file, the start andend position, the frequency of acquisition, as well as the processing history. The lower section shows a final raw data plot. In this case polarity, statics and topography corrections have been applied, three files have been merged and allthis information is annotated properly for future reference. A basic time gain has been applied to make the display.

(a)

(b)

Figure: 8-2 a) Example of data processing documentation, b) example data set from an archaeological investiga-tion whose history is described in 7.2. The data were assembled from three data sets and have had tomography,

polarity and static correction applied. An SEC time gain is applied in the plotting.

8/20/2019 GPR Annan 2003



159

To date, multifold GPR data acquisition is not done extensively. If it becomes more common, this step of processingis the one where re-organization and sorting of multifold coverage data would be handled. The steps are essentiallythe same as with multifold seismic data (Fisher et al, 1992a).

8.2 BASIC PROCESSING

Basic data processing addresses some of the fundamental manipulations applied to data to make a more acceptable product for initial interpretation and data evaluation. In most instances this type of processing is already applied inreal-time to generate the real-time display. The advantage of post survey processing is that the basic processing can be done more systematically and non-causal operators to remove or enhance certain features can be applied.

8.2.1 DEWOW

The initial basic processing step is usually temporal filtering to remove very low frequency components from thedata. This step is frequently referred to as 'de-wowing' the data. Very low frequency components of the data are asso-ciated with either inductive phenomena or possible instrumentation dynamic range limitations. This process has his-torically been done using analog filters in hardware but with the advent of true digital acquisition this has also become a data processing issue (Gerlitz et al, 1993).

8.2.2 TIME GAIN

The next step of basic processing is usually to select a time gain for the data set. Time gain has historically been verysubjective and also very much display device dependent.

Radar signals are very rapidly attenuated as they propagate into the ground. Signals from great depth are very smalland display of this information at the same time as signals from a shallower depths is difficult. When the amplitudeof display is optimal for shallow depth signals, events from greater depths may be invisible or indiscernible. Equaliz-ing amplitudes by applying some sort of time dependent gain function compensates for the rapid fall off in radar sig-nals from deeper depths. This is referred to as "time gain".

Figure 8-3 indicates the general nature of the amplitude of radar signals versus time. On the left there is a idealizedearth which has horizons at equal depth intervals which all have the same reflectivity. On the right is an idealizeddepiction of the amplitude of the returns from each horizon as a function of time.

Figure: 8-3 Layered earth model of equal reflectivity horizons and impulse response with envelope of reflectionamplitude depicted.

8/20/2019 GPR Annan 2003



160

A major problem with GPR data is that the attenuation in the ground can be highly variable. One can have a lowattenuation environment where exploration depths of 10's of meters can be readily achieved. In other situations atten-uation can be quite high and depths of only 1 to 2 meters can be penetrated. The behavior of the radar record versustime is depicted in Figure 8-4 for the low and high attenuation extremes.

Figure: 8-4 Equal reflectivity layered earth response in lower and higher attenuation media.

A simplified way of viewing the amplitude of the signals versus time is shown in Figure 8-5. A spherical wavespreading into the ground will fall off inversely as the distance of the waveform into the ground and will also die outexponential depending on the conductivity losses in the material. Assuming a constant velocity earth it is possible totransform depth into time and hence derive an amplitude of the signal versus time for the spherical wave reflectingfrom the simplified equal amplitude reflection layered earth.

AMPLITUDE VERSUS DEPTH

A(d) = exp (-2 a d) / 2d

AMPLITUDE VERSUS TIME

A(t)= exp ( -a*v*t) / (v*t)

a = attenuation d = depth v = velocity t = time

Figure: 8-5 Concept of translation amplitude fall of versus depth to fall off versus time.

The concept of time varying gain is simple. One applies a gain to the data which increases in time after the transmit pulse. The rise of the gain function is tailored to accommodate the drop off and signal amplitude versus time. Theconcept is depicted in Figure 8-6.

A°

A°

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



161

Figure: 8-6 Concept of time varying gain where signal amplification varies with time.

In order to put the variation in amplitude versus time into perspective, the average amplitude on a radar section wascomputed. The amplitude versus time is shown in Figure 8-7. In can be seen that the signal drops several orders ofmagnitude from time zero out to a time of about 300 or 400 ns at which point the signals drop back down to the noiselevel. An indication of the noise levels can be estimated from the data before time zero on the plot. In this case thenoise level is about 10 to 15 microvolts.

Figure: 8-7 Composition display of radar signal versus delay time for a radar section.

Systematic ways of selecting gain are available. Generally the first step in selecting gain is to examine the amplitudeversus time fall-off of the data. This can be done trace by trace or on a section basis. Generally, working on a sectionis the most useful way of approaching the problem but again there are degrees of variability in this. Figure 8-8 7.8shows an average amplitude versus time plot for high pass (dewow) filtered but otherwise raw data set shown in Fig-ure 8-2. The key point to note is that the radar signal amplitude falls off with time in a fairly systematic fashion. Ifone wants to display these data on a display device with a dynamic range of 20 dB (i.e., a factor of 10 between thesmallest and largest signal) then the 60 dB dynamic range of the real signal has to compressed into a dynamic rangeof 20 dB of the display device. Figure 8-9 shows the average signal amplitude versus time after the time gain used for

8/20/2019 GPR Annan 2003



162

the plot in Figure 8-2 has been applied.

Figure: 8-8 The average amplitude of the radar signal versus time for the data set presented in Figure 8-3 prior totime gain.

Figure: 8-9 The average signal amplitude of the radar signal versus time for the data set presented in Figure 8-2bafter the application of an SEC time gain.

There are a variety of ways of applying gain to radar data. The issue is whether or not one wants to maintain somesort of amplitude fidelity or whether one only wants to display all of the signals in the data. For stratigraphic horizoncontinuity sometimes showing all the information irrespective of amplitude fidelity is important. In this case a con-tinuously adaptive gain such as AGC (automatic gain control) is used. With AGC gain, each data trace is processedsuch that the average signal is computed over a time window and then the data point at the centre of the window isamplified (or attenuated) by the ratio of the desired output value to the average signal amplitude.

Physical phenomenon based systematic gains, such as spherical and exponential compensation (SEC) gain, attempt toemulate the variation of signal amplitude as it propagates in the ground. The data displayed in Figure 8-2 have had anSEC gain applied and comparing Figure 8-8 and Figure 8-9 shows the degree to which the time gain has compressedthe signal dynamic range.

Examples of SEC and AGC gain on a short set of radar traces are shown in Figure 8-10 and Figure 8-11. These are practical gain implementations which limit the maximum amount of gain that can be applied.

A variety of other gains can be applied. Gain should be selected based on some a priori physical model, not at theuser whim. The objective should be to modify the data while retaining its full utility without introducing artifacts.

8/20/2019 GPR Annan 2003



163

Figure: 8-10 Illustration of raw data traces, the spherical and exponential compensation (SEC) gain versus time

and the resulting gained traces.

Figure: 8-11 Illustration of raw data traces, the automatic gain compensation (AGC) gain versus time and theresulting gained traces.

8.2.3 TEMPORAL AND SPATIAL FILTERING

Temporal and spatial filtering are often the next stage of basic processing. Filtering can be applied before or aftertime gain as long as the effect of the gain is understood since time gain is a non-linear process. Temporal filteringmeans filtering along the time axis of the data set. A whole host of different types of temporal filtering may beapplied from bandpass filtering using fast Fourier transforms (FFT) through to various types of linear and non-lineartime domain convolution filter operators. Figure 8-12 shows the amplitude spectra for the data set shown in Figure 8-2 before and after an SEC gain function has been applied. Note the change in spectra induced by the time gain whichis a non-linear operator. In both cases the average amplitude spectrum for the whole section has been generated.From the spectra it can been seen that the majority of the energy is limited to a finite bandwidth and appropriate useof band limiting filtering may improve signal-to-noise without significantly altering the data. A low pass filter withan 50 MHz corner using an FFT approach generates the section shown in Figure 8-13a while Figure 8-13 b shows theresult of a 80 to 120 MHz bandpass filter.

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



164

Figure: 8-12 Average time amplitude spectrum shown in Figure 8-3 prior to time gain. b) Average amplitude spec-trum for data shown in Figure 8-3 after application of an SEC time gain.

(a) (b)

Figure: 8-13 a) Data set shown in Figure 8-2b after low pass zero phase FFT filter with a cosine taper between 50 MHz and 100 MHz. b) Data set shown in Figure 8-3 after 80 to 120 MHz bandpass zero phase FFT filter.

Similar filtering operations can be applied in the spatial domain. The 3 radar sections shown in Figure 8-14 present

the original radar section plus an example of high pass and low pass spatial filtering. The high pass spatial filterretains dipping events and suppresses flat lying events. The low pass filter has the opposite effect in that it enhancesflat lying events and suppresses dipping responses.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



165

(a)

(b)

(c)

Figure: 8-14 a) Example of time gained data in fluvial sand environment. b) Illustration of spatial low pass filteringas applied to data shown in a. Note enhancement of flat lying reflectors. c) Illustration of high pass filter applied to

data shown in b. Note enhancement of dipping events and diffraction details.

Median and alpha mean trim filters offer powerful data "clean-up" filters for noise spikes. These filters can beapplied in both the time or space domain. These filters may be applied before and after time gain but are usually mostuseful if applied before time gain and any type of data adaptive filtering. An example of spike removal from a data setusing a 3-point temporal filter followed by trace spatial median filter is illustrated by the before and after sections in

Figure 8-15.

8/20/2019 GPR Annan 2003



166

(a)

(b)

(c)

Figure: 8-15 a)An original data set with a large amount of spurious localized noise events. b) The data shown in a)after a 3-point temporal median filter. c) The data set shown in Figure 8-15b after a 3-point spatial median filter.

8.2.4 DISCUSSION

The examples cited in this discussion are by no means exhaustive and a wide variety of simple data manipulations areavailable to enhance basic aspects of GPR data. As evident from the preceding, the order in which these processesare applied can also be varied producing different results because of the non-linear nature of some steps. As indicatedearlier, the bias in this discussion is to single fold reflection data.

Basic data processing should leave the data sets reasonably intact. In other words, the processing should not be suchthat it radically distorts the information from that which was collected. Again the degree of distortion is subjective

and obviously excessive bandpass f iltering can drastically alter a data set as evidenced by the spatial high and low pass filtering shown in Figure 8-14. Normally minor changes to the overall data set occur if simple basic processingsteps are applied intelligently.

8.3 ADVANCED DATA PROCESSING

Advanced data processing addresses the types of processing which require a certain amount of operator bias to beapplied and which will result in data which are significantly different from the raw information which were input tothe processing. Such processes include well-known seismic processing operations such as trace attribute analysis, FK

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



167

filtering, selective muting, normal move out correction, dip filtering, deconvolution, and velocity semblance analysisas well as more GPR specific operations such as background removal, multiple frequency antenna mixing and polar-ization mixing (Tillard & Dubois, 1992).

8.3.1 ATTRIBUTES

Dealing with all these topics is impossible in the space available in this paper but two examples will be provided.

One topic of interest is the use trace attribute analysis common in seismic processing (White, 1991). In this case atime series is decomposed using Hilbert transform and minimum phase assumptions into a real and imaginary timeseries from which the envelope and the frequency can be estimated at every point along the trace. Figure 8-16 showsthe envelope of the data set originally shown inFigure 8-2a. Figure 8-17 shows the instantaneous frequency computed for the same data set.

Figure: 8-16 The amplitude envelope for data shown in Figure 8-2b. The envelope more accurately reflects the GPRresolution.

Figure: 8-17 The instantaneous frequency for the data shown in Figure 8-2b. The instantaneous frequency tends toreflect the spatial scale of the radar targets.

The benefit of the envelope display is that it reflects the resolution of the data. One tends to get a false sense of reso-lution because of the oscillatory nature of the radar pulse. In fact the bandwidth or envelope of the pulse is whatdetermines resolution not the time between zero crossings (which reflects the dominate frequency in the data). Forthis reason the envelope is extremely useful for generating more simple presentations which are representative of thedata spatial resolution.

The instantaneous frequency can be a good indicator of certain types of features which selectively respond to thespectrum of the incident signal. An example of this is a thin layer in a stratigraphic sequence. If the layer is close tothe thickness where the pulse spectrum is at the tuning frequency of the layer then an enhancement in the response of

8/20/2019 GPR Annan 2003



168

that particular frequency will occur. Instantaneous frequency gives a rough feel for the texture on spatial scale of theGPR signal scattering sources.

8.3.2 DECONVOLUTION

Examples of deconvolution and other types of filtering are given by Todoeschuck (1990) and Turner (1992) . It is

important to note that deconvolution ("decon") of GPR data is not straight forward and has seldom yielded a greatdeal of benefit. Part of the reason for this is that the radar pulse is often as short and compressed as can be achievedfor the given bandwidth and signal-to-noise conditions. Another important factor is that some of the more standarddeconvolution procedures have underlying assumptions required for wavelet estimation such as minimum phase andstationarity which are not always appropriate for GPR data. The rapid decrease in GPR signal amplitude means thatdecon artifacts may mask weaker deeper events if time gain is not applied before decon and the non-linear nature oftime gain may substantially alter wavelet character if gain is applied before decon. As a result decon can be both dif-ficult to apply systematically and exhibit little enhancement in resolution. Instances where deconvolution has proven beneficial occurred when extraneous reverberation or possibly system reverberation have been involved. Deconvolu-tion can then provide substantial pulse compression benefits.

8.3.3 BACKGROUND SUBTRACTION

One of the most common operations specifically applied to GPR data is the use of background removal. Most oftenthis takes the form of a high pass filter or an average trace removal. More sophisticated approaches which yield sim-ilar results are orthogonal trace decomposition (Friere & Ulrych, 1988).

Average trace removal is a form of spatial filtering. In some situations where transmitter reverberation and time syn-chronous system artifacts appear it is very effective in allowing subtle weaker signals which are lost to become visible in a processed section. Figure 8-18 illustrates this concept with before and after sections where average traceremoval has been applied. In this case, the target is shallow and the ground response is masked by the transmit pulse.

This type of processing is often necessary, but should not be used routinely. If data always exhibit the need for this processing serious equipment flaws are present in acquisition.

(a)

(b)

Figure: 8-18 a) Example of initial data set over a buried pipe. b) Data set shown in a) after the average trace forthe whole section has been subtracted. The hyperbolic response is clearly visible as are the gently sloping edges of

the excavation.

8/20/2019 GPR Annan 2003



169

8.3.4 VELOCITY ANALYSIS FROM CMP

Another radar data processing is to analyze velocity sounding data to extract velocity versus depth functions. Thiscan be done by picking events and using T-2 - X-2 analysis or a variety of other methods. Figure 8-19 shows how aCMP data set has been stacked using a move out correction where the constant velocity can be applied. The resultantstacked data shows the velocities which the data add up most coherently.

Figure: 8-19 CMP Data.

Much more sophisticated versions of this algorithm are available in many of the seismic processing packages. Quiteoften they are referred to as assemblance analysis routines. Example results are given in Figure 8-20.

Figure: 8-20 Velocity versus depth.

8.3.5 DISCUSSION

The important feature of advanced processing is that it focuses on making weaker signals visible, enhances specificcomponents of the data for an interpretation requirement, or derives quantitative information such as velocity andattenuation versus depth from the data. The advanced processing methods also have the potential of introducing arti-facts in areas where there are no responses in the ground giving rise to specious interpretations. Considerable userinsight is required when this type of processing is applied so that undue emphasis is not placed on artifacts induced inthe response by the processing.

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



170

8.4 VISUAL/INTERPRETATION PROCESSING

Processing in this context is usually conducted when a good deal of information about a site is available and a needfor the processing has been defined which achieves a final objective. Processing in this class will often result in datawhich are totally changed from the original data set. Furthermore the type of processing is usually subjective in thesense that there is some a priori bias in the processor's mind as to what end product is desired.

processing in this class includes topics such as migration using various types of algorithm, event picking, subjectivegain enhancement and amplitude analysis. All of these require completion of the previously mentioned processingsteps and availability of corollary control information.

8.4.1 MIGRATION

For example, migration is extremely useful and reconstructs the radar image in a form which is probably a better rep-resentation of the ground (Fisher et al, 1992b) Unfortunately migration requires a good knowledge of the velocitystructure in the ground and a processor who is aware of the artifacts that migration introduces into the section. As aresult it is dangerous in the hands of a novice but powerful in the hands of a processor who has acquired the ability touse it effectively and recognizes the limitations. Migration is often an interactive process as background velocity isadjusted to optimize the migrated result.

Figure 8-21 shows an example of data from a site before and after migration (Brewster, 1993). In this case the migra-tion has been used to remove defractions from vertical sheet pile walls at the edge of the site. These metal sheets

were driven down into the soil to act as a containment barrier. The reflections and scattering from these metal wallsare visible across most of the cell and tend to mask some of the other features in the data set. Migration very conve-niently removes the diffractions by placing these events back in the correct spatial location. While an enhancement isquite visible in the data, it is not easy to assess what distortion, if any, have been generated.

(a)

(b)

Figure: 8-21 a) Example of GPR data from a contaminated DNAPL spill. Note the dipping events which slopedown from each end of the section associated with scattering from sheet pile containment walls driven into the

ground. b) Data in a) after F-K migration. Note that the wall events have been migrated out of this section. The strong reflection between 2 and 6 m position at times of 25 to 40 ns is the return from a DNAPL pool.

8/20/2019 GPR Annan 2003



171

The key to success in the migration process rests in having a good knowledge of the velocity section and quite oftenwith single fold GPR data this information is not available. Migration is then very much at the discretion of the pro-cessor rather than controlled by the site conditions. Iterative migration by varying velocity structure to make the mostacceptable (to the processor) result is common.

8.4.2 EVENT PICKING

Another way of manipulating data is to pick events and display the reduced event data set. This is a very useful thing because it simplifies the data set and pulls out only those features which are deemed important. The important pointto note is the very subjective and application dependent nature of the processing. Figure 8-22 shows the result of picking both amplitudes and arrival time data from a 2D grid of which Figure 8-21 shows one section to create a tar-get depth versus position plot over an area. This information was subsequently used to estimate the volume of aDNAPL contaminant in the ground.

Pool 1 Topography, at 36 hours

Figure: 8-22 Depth to top of the DNAPL pool shown in Figure 7-21 based on an event picking grid of survey lines. Figure 7-21 shows the radar cross section of 5 east.

Visual/interpretation processing are most useful when one wants to move to 3D visualization of data versus space andtime. For exam ple, Figure 8-23 from Brewster (1993) shows a 3D view of DNAPL contamination pools for whichdata shown in Figure 8-21 and Figure 8-22 were prior stages of processing. While raw data can be displayed thisway, most often the essential information is enhanced and the rest of the information suppressed. Frequently the end product is the most easy to present to non-technical users. This is perhaps the most dangerous area of data processingin that the visualization may over simplify the problem and give a false impression of real conditions.

8/20/2019 GPR Annan 2003



172

Figure: 8-23 Simplified time and 3D spatial visualization of DNAPL pools based on procedures as illustrated by Figures 7-21 and 7-22.

8.4.3 VOLUME VISUALIZATION

Affordable access to 3D visualization tools and the rapid advance in computer technology are opening these tools toeveryday GPR analysis. The degree of preprocessing needed to make these displays effective varies and encompasses all of the topics discussed here and more. The power of collecting reflection survey data in a tight grid anddumping it into 3D display software is enormous. Trends and subtle hard to correlate events become very visiblewhen displays are animated.

In general, these displays lose their impact when printed in hard copy form. To indicate the types of display availablea number of examples are given in Figure 8-24; this includes chair cuts, slices and variable capacity.

8/20/2019 GPR Annan 2003



173

Figure: 8-24

8.5 DISCUSSION

The preceding has provided a brief overview of the steps of data processing which are entailed in GPR analysis. Ashas been previously stated, the amount of processing can vary from none through to quite sophisticated analysis andmanipulation.

As some of the simple processing steps presented here have vividly illustrated, processing can markedly alter the

appearance of a data set. As a result it is very important that a processing trail be maintained for audit purposes.Since survey results in engineering and environmental projects may be used as evidence in legal proceedings, it isimperative that the means of creating a processed data set be reproducible. Figure 8-2a illustrates how the historicalgenesis of a processed data set is maintained in a commercial software package.

All of the processing methods discussed and display methods illustrated can be carried out on readily available per-sonal computers with relatively low cost commercial software. Anyone who wishes to process GPR data can do sowith a modest capital outlay.

Perhaps the most important fact that should be kept in mind is that there must be an end objective. For any GPR sur-vey, a value of some sort is always attached to the survey objective. (Otherwise why do it in the first place??) How-

8/20/2019 GPR Annan 2003



174

ever this value is defined, one must remember that processing entails time, resources and money. While people donot always recognize it, the biggest single cost is the human time entailed in data manipulation, data organization andthought. Usually this requires a reasonably well trained person whose time is not inexpensive. While computer costsmust be considered, man power costs frequently exceed equipment costs.

Figure 8-25 shows a simplified chart of the above concepts. The vertical axis shows a value or a financial cost impli-cation. The horizontal axis shows a measure of time elapsed during processing. As the degree of processing

increases, the amount of time and hence costs increase.

Figure: 8-25 Illustration of how the value of GPR survey is fixed while costs climb as time passes as will ensure ifiterative and uncontrolled data processing occur.

t is important that this type of analysis be carried out before one does extensive GPR processing. By taking the sur-vey value and dividing it by cost, one obtains a benefit-to-cost ratio and processing should not be carried on beyondthe point where the benefit-to-cost ratio is one as indicated inFigure 8-26.

Figure: 8-26 As time and processing increase, the benefit-to-cost ratio decreases. When the ratio approaches unity,the benefit of further processing and analysis becomes dubious.

8/20/2019 GPR Annan 2003



175

Unlike some applications of geophysical methods such as petroleum or mineral exploration where quite large finan-cial rewards exist at the end of the successful project, there are smaller and quite often much more poorly definedrewards at the end of engineering and environmental geophysics projects. Since this is the area where GPR finds itsuse most often, the benefit-to-cost ratio is one which has to be kept in mind and frequently approaches unity early inthe data processing scheme.

While research institutions and government agencies with unconstrained man-power and resource availability can

carry out much more advanced processing, many commercial applications often have benefit-to-cost ratios such thatthe processing is terminated at the field acquisition step. In other words, no processing is done!!

8.6 FINAL WORDS

The preceding has been an attempt to provide an overview of the current state-of-practice of ground penetrating radardata processing. While less than an exhaustive treatise, the important points that should be noted are the following:

a) GPR data processing is within the financial and technical reach of anyone who wishes to process data, not just the few who have extensive high power computing facilities andsoftware development resources.

b) Many of the vast array of reflection seismic processing techniques can used directly onGPR data.

c) GPR data are not totally identical to seismic data and there are GPR specific processingtechniques.

d) An audit trail should be an integral part of any data processing program.

e) Data processing has to have a cost benefit and, for many GPR projects, processing is costlimited rather than technology or methodology limited.

The processing of GPR data is still truly in its infancy. The flood gates are opening to GPR world wide and there isan explosive growth in GPR awareness. These factors will feed a growth in GPR processing because more and betterGPR results will be obtained, advances in processing and presentation will occur, processing costs will decrease anda higher value will be placed on GPR results. The cost benefits of processing will become easier to establish andhence processing will be more readily justifiable and more frequently used.

8/20/2019 GPR Annan 2003



176

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 9-Interpretation, Concepts & Pitfalls

177

9 INTERPRETATION, CONCEPTS & PITFALLS

Interpretation is a very objective dependent activity and is also very subjective (in that it depends greatly on the indi-

vidual doing the interpretation). Since time has not permitted an extensive dissertation on this subject, this chapter is

limited to pointing out some aspects and pitfalls. A variety of case history data examples follow.

The following are the critical aspects of data interpretation.

1. Have a good understanding of the survey objective.

2. Develop a model of the geological setting or application structure.

3. Organize the data into systematic sets which are easily correlated with site maps.

4. Establish an estimate of velocity and attenuation (or depth of exploration achieved).

5. Develop a scenario for the radar response expected (i.e. a slowly varying continuous event

for a major geologic horizon, a spatially limited diffraction hyperbola for a pipe). If possi-

ble process data to enhance the type of response expected.

6. Correlate radar data with geologic control and ground truth such as drilling.

7. Plot radar anomalies locations on a plan map to permit line to line correlation.

8. Plan drilling or follow up control work.

9. Review data after additional control becomes available.

The following are small vignettes addressing specific problems.

9.1 Gradational Interfaces

Many GPR problems require detection of the response of interfaces which are not sharp and clean cut. Natural inter-

faces are often blurred by the natural depositional sequence involved in the creation of the stratigraphy. Many man-

made interfaces can also be indistinct as construction practice may lead to mixing of materials.

The water table example is a classic example of a gradational boundary. The water table has one of the largest prop-

erty contrasts in nature yet invariably it may not yield a strong radar response, particularly at higher frequencies.

The gradational nature of the water table interface is depicted in Figure 9-1. Water is distributed in a gradational

fashion in the pore space due to capillary suction.

8/20/2019 GPR Annan 2003


9-Interpretation, Concepts & Pitfalls GPR Principles, Procedures & Applications

178

Figure 9-1: Schematic illustration of how water content varies with depth. The change from residual water content to saturation

is note sharp but gradual. Transition width depends on soil texture.

Figure 9-2 depicts a simple model of a gradational boundary and a sharp boundary. The transition in the gradational

case spans a width of 0.5 meters which is typical of the transition width in the Camp Borden sands in the examplesection presented in Figure 9-5. The synthetic GPR responses for a 100 MHz radar system are shown in Figure 9-3.

Figure 9-2: Two model profiles of permittivity versus depth - one abrupt and one with graded permittivity over a 0.5 width.

From the simple modelling results, it is apparent that introducing a gradational transition substantially reduces thewater table reflection amplitude. It also smears the reflection out with lower frequency content; the gradational

boundary acts as a low pass filter on the GPR pulse.

The synthetic data are also plotted in cross section form in Figure 9-4. The contrast between the sharp boundary and

the gradational boundary response is very apparent. The blurred response of the gradational interface is quite similar

to that observed in the radar sections shown in Figure 9-5.

8/20/2019 GPR Annan 2003



179

Figure 9-3: Results of synthetic radar calculation for two cases shown in Figure 9-2.

Figure 9-4: Gray scale displays of the sharp and gradational responses on a comparable scale to the GPR sections in Figure 9-5.

The gradational response has reflections width similar to the water table reflection.

Figure 9-5: The figure on the left shows a 100 MHz pulseEKKO 100 section of water table test line collected in step mode at 20

cm station intervals. Antennas axes were aligned perpendicular to the survey line. The figure on the right is the same as the left

only this shows antennas were axes oriented parallel to line direction and antennas moved continuously resulting in less accurate

spatial positioning. Rubber banding was used to covert data to equal spatial positioning prior to plotting.

8/20/2019 GPR Annan 2003



180

In summary, many interfaces are not as sharp and clean as the ideal case. GPR responses will tend to be weaker and

more blurred than the ideal. These are important concepts for GPR users to keep in mind.

9.2 Velocity Determination Using Hyperbolic Fitting

The full utility of GPR data requires knowledge of how fast the signals travel in the material under investigation.Several techniques have been used such as CMP (common mid-point), WARR (wide angle reflection and refraction),

known-depth-target, hyperbolic fitting to a local target and diffraction tail matching.

All of these techniques require GPR measurements along a traverse where the geometry is varying in controlled fash-

ion. In other words, the distance to a target varies in such that estimations of velocity can be extracted.

For pipe and cable location of rebar and conduit location, line-like features are localized targets if the GPR system

traverses perpendicular to the feature alignment. To estimate velocity, the path length to the object must vary. Figure

9-7 illustrate this using a linear pipe or cable as an example. In order to extract velocity information, the radar sys-

tem must be moved perpendicular to the axis of the pipe or cable. The long-axis direction is commonly called the

“strike direction” or “strike” for short. If a GPR traverses perpendicular to the strike, the distance varies from the

radar system to the pipe in a systematic fashion. Traversing parallel to the pipe strike yields no change in the distance

of the pipe and hence, a flat non-changing event on the GPR record. Figure 9-8 shows these two extremes using realdata from a clay drainage pipe in a farm field.

Figure 9-6: Cross section through area with GPR traverse perpendicular to pipe or cable strike direction.

Figure 9-7: Plan view looking down or ground from above. Traverse 1 is perpendicular to strike and is optimal for velocity

determination. Traverse 2 is at an oblique angle and 3 is parallel to the pipe strike axis. Data from traverses 2 and 3 are not suit-

able for determining velocity.

PIPE

1

2

3

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



181

Figure 9-8: GPR data over clay drainage pipe perpendicular to pipe direction (line 1 in Figure 9-7 )

Figure 9-9: GPR data over a clay drainage pipe parallel to pipe direction (line 3 in Figure 9-7 ).

Figure 9-10: Relationship between GPR position (x), object depth (d) and travel time (T). T 0 is travel time when GPR is directly

over the object .

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



182

GPR cross-sections display signal amplitude versus position (normally on the horizontal axis denoted as x) and time

(which is normally the vertical axis denoted as T). A local target has a travel time versus position as depicted in Fig-

ure 9-10. The mathematical form is a hyperbolic shape (inverted U on a GPR section) relating spatial position (x) to

travel time (T). Figure 9-11 shows the response in a GPR cross-section as the target depth is varied. Figure 9-12

illustrates the behavior as the velocity is changed.

Figure 9-11: Schematic variations in GPR response when object depths varied for constant velocity.

Figure 9-12: Schematic variations in GPR response when velocity is varied for a fixed object depth.

Figure 9-13: Example of shape fitting to a target response on the screen in the field. This feature is standard on Noggin Smart

Systems and Conquest.

8/20/2019 GPR Annan 2003



183

A handy interpretation aid is to visually fit a model hyperbolic shape to the GPR data as illustrated in Figure 9-13.

Placing the top of the model (triangle point) over the apex (top of inverted U) in the data section selects T0. Adjusting

the model shape to match the data yields an estimate of the velocity, v. Combining v and T0 yields an estimate of the

depth to the top of the target.

This tool is routinely employed in many of Sensors & Software Inc.’s Smart Systems. Users must be cautious since

the velocity estimate can be erroneous if the traverse is not perpendicular to a linear feature strike. Good field prac-tice entails several traverses over an object. Only use the hyperbolic fitting on the traverse that gives the most steep

slope to the arms of the inverted U. This approach assures getting the most correct velocity. A traverse not perpen-

dicular to a strike will be always yield a velocity higher than the true velocity object depths will appear deeper than

reality.

9.3 Polarity

The polarity of a GPR signal can be very helpful in interpretation of data. In this section, we will provide some basic

insight into polarity, what causes it and how you recognize it in data.

The first thing is to define what we mean by polarity. A GPR signal normally reduces to a wavelet in time which has

three half cycles such as shown in Figure 9-14. This up-down-up or down-up-down nature of the signal is commonand is the result of the radiation characteristics of a small dipole antenna. The relative amplitude of the half cycles are

1, -2, 1 or –1, +2 or -1. The resultant signal has no DC or average value.

The user must develop a systematic convention for antenna deployment in order that the sign convention is consistent

for a given survey application. There is no particular standard for what is defined as positive and negative. Normally

at Sensors & Software Inc. we define the polarity of the signal based on the sign of the voltage of the first half cycle.

A positive voltage in the first half cycle means a positive pulse while a negative voltage during the first half cycle

means a negative pulse or polarity.

Sometimes it is easier to describe a positive wavelet according to Sensors & Software Inc.’s sign convention as a sig-

nal that looks like an “M”. A negative polarity wavelet would be one that looks like a “W”. These concepts are

depicted in Figure 9-14.

Figure 9-14: The typical GPR wavelet is an “M” or “W” shaped pulse as depicted here.

When a GPR measurement is made, there are signals which travel directly from the transmitter to the receiver as well

as reflections coming from the subsurface. Figure 9-15 depicts the basic concepts of a simple measurement showing

the direct air wave, the direct ground wave and a subsurface reflection. In this case we use a positive transmit pulse

(having an “M” shaped wavelet). The direct airwave will have a positive “M” shaped wavelet whereas the direct

ground wave will have a negative or “W” shaped wavelet. The polarity of the reflection from the subsurface horizon,

however, can either be positive or negative (“W” or “M” shaped).

8/20/2019 GPR Annan 2003



184

Figure 9-15: When GPR measurements are made, the transmitter and receiver are placed on the material to be probed. The min-

imal number of events which will normally be observed are 3 which are the direct wave A, the direct ground wave G, and a

reflected signal R. Two idealized radar traces are shown here which show the polarity of the various wavelets associated with the

arrivals. Two traces are shown because the reflected signal could have either a positive or a negative polarity.

The polarity of the reflected signal is dictated by the nature of the change in the electrical properties which cause the

reflection. Figure 9-16 depicts the details of reflection from a target interface. Reflection is caused by a change in the

electromagnetic impedance. A higher impedance target yields a positive reflection coefficient whereas a lower

impedance target yields a negative reflection coefficient. In GPR, metal is a very low impedance (essentially 0)

material and will always give rise to a negative reflection wavelet. On the other hand, a void in soil or rock represents

a high impedance material and will give rise to a positive reflection wavelet.

Figure 9-16: When a signal is incident on an interface, the reflected signal is dictated by the change in impedance on the inter-

face.

8/20/2019 GPR Annan 2003



185

An example data set is shown in Figure 9-17. These GPR measurements were made on a concrete structure which

contained metal rebar and an air filled duct. The polarity of the signs of the wavelets from the reflections are clearly

in accord with the above discussion. In the particular sections shown here, positive signals are shown as black, nega-

tive signals are shown as white and background level signal (which are zero amplitude) are shown as gray.

Figure 9-17: An example of GPR data from concrete data. Positive wavelets are black-white-black bands, negative wavelets are

white-black-white bands and the background gray represent zero signal level. In this case the radar profile passed over a metal

reinforcing bar and an air filled non-metallic conduit in the concrete.

9.4 Airwave Events

‘Airwaves’ is the name given to events on GPR records which are associated with energy that leaks into the air and

gets reflected back into the GPR receiver. The source of these airwave reflections can be any above ground objects

such as telephone poles, overhead wires, walls, vehicles, etc. (see Figure 9-18).

Figure 9-18: Radar energy that leaks into the air can reflect from surface objects in the GPR survey area. These ‘airwave’ reflec-

tions can look like subsurface targets in the GPR record.

GPR users tend to think that all GPR responses come from within the ground. It is very natural to think that GPR sys-

tems only look downwards and implicitly believe any event on a record must be from the subsurface. In fact, because

the ground absorbs energy so rapidly whereas air does not, airwave events will often appear on GPR records.

8/20/2019 GPR Annan 2003



186

Airborne signals cause problems in two different ways:

• the airwave response may mask weaker subsurface signals and make them difficult or impossible to

see or interpret,

• airwaves may be falsely interpreted as a buried object when in fact the object is above the ground.

The GPR data set shown in Figure 9-19 is an example of how a false interpretation could occur. This particular data

set was acquired along a road profile. The objective of the survey was to see whether there were karst (void) features

underneath the roadway which could pose hazards to vehicular traffic. Several hyperbolic responses typifying the

presence of localized features were apparent on the record.

Figure 9-19: Airwave reflections from trees in the GPR survey area. All the reflectors below 100 ns in the data are caused by sur-

face objects.

To the uninitiated, these ‘hyperbolic’ features can easily be misconstrued as coming from localized cavities beneath

the roadway. An experienced GPR operator should routinely question the source of all signals particularly when the

amplitude versus time response indicates that signals in the ground are decaying at a very rapid rate. The amplitude

of the hyperbolic responses in Figure 9-19 are anomalously large given the attenuation in non-anomalous areas of thedata.

A second reason that these features can be identified as airwave events is the slope of hyperbolic tails. These events

have an airwave slope which indicates that the source of the signal is in the air.

Indicators of airwaves are:

• use your eyes to look for potential sources of airwave in your survey areas;

• when profiling, always run some profile lines towards or away from a surface object to evaluate

whether or not the object is scattering energy. The response from the object will show up on the

GPR record as a straight line event which slopes at the airwave velocity (see Figure 9-20, Figure 9-

21 and Figure 9-22 lines B – B1 and C – C1);

• check diffraction tails of hyperbolic events for slopes close to the airwave speed; and

• check the frequency content of events as airwave energy often has higher frequency content than

subsurface responses.

8/20/2019 GPR Annan 2003



187

Figure 9-20: A figure showing the orientation of 3 GPR survey lines (A, B and C) in an area where a building and a tree are

present. The figure also shows how during data acquisition of LIne A, radar energy leaks out into the air, reflections from the

buildings and the tree and returns to the receiver. Although not depicted the same leakage occurs on Lines A, B and C are shown

in Figure 9-21.

Figure 9-21: Airwave responses from LIne A in Figure 9-20. As the GPR survey line is run from A to A’ it approaches and passes

the building and then approaches and passes the tree. The sloped responses occur as the distance to a surface object decreases as

it approached or increases as it is passed. When the GPR is run parallel to a long linear surface object like a building, the air-

wave response is a flat event. When the GPR is run beside a point target like a tree the airwave response is the classic hyperbolic

shape.

Figure 9-22: Airwave response from Line B in Figure 9-20. As the GPR survey line is run from B to B’ the building gradually

gets further away. Consequently, the building creates a straight reflector starting early in time at position B and later in time at

B’. The slope of this is 1/c where c is the speed of light.

8/20/2019 GPR Annan 2003



188

Figure 9-23: Airwave response from Line c in Figure 9-20. As the GPR survey line is run from C to C’ the tree gradually gets fur-

ther away. Consequently, the tree creates a straight reflector starting early in time at position C and later in time at C’. The slope

of this line is 1/c where c is the speed of light.

Removing or eliminating air launched energy is difficult. Sometimes it can be minimized by optimizing antenna and/

or survey line orientation. Airwave tends to get launched as vertically polarized fields off the ends of the dipolar

GPR antennas traditionally used. Vertical features such as pipes, poles and trees make good targets. Changing the

antenna orientation is only practical in some types of surveys and not all situations.

Digital processing may filter out the airwave features. For survey lines with airwave targets are at or beyond the end

of the line, the airwave event will have a fixed slope (Figure 9-22 and Figure 9-23) allowing digital dip filters to

remove the airwave energy and leave all the rest of the data intact.

In other situations the difference in frequency content between ground events and airwave events can be used to

advantage. Low pass temporal filtering can suppress airwaves and enhance ground responses.

In some situations GPR just cannot be used because the airwave events dominate. This is a fact of life and must beaccepted as one of the limitations of GPR.

Avoid the fallacy that shielded antennas will eliminate airwaves. Shielding of GPR antennas is never fully effective.

Shielding can reduce airwave events (by a factor of two to even ten) but it never fully eliminates their presence.

“Shielded” antenna data should always be treated as suspect when it comes to airwaves.

9.5 X Marks the Spot

For many years GPR records with rather unique X character have cropped up. Figure 9-24 shows 2 examples of such

records.

These examples show several partial X’s. This characteristic X is caused by GPR measurements made inclose prox-

imity to a long length of wire such as fence or cable.

The concepts underlying the cause of the response are depicted in Figure 9-25. When a GPR system is placed near a

long metal wire, some of the energy couples from the transmitter onto the wire. The energy then travels along the

wire in both direction until it reaches the end of the wire and then bounces back.

As one moves the radar system along beside the wire, the point at which the energy jumps on to the cable or wire

moves the reflection from the ends of the wire move in time on the GPR record.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



189

Figure 9-25 explains how the geometrical changes the delay times for the signals traveling along the wire. Generally

the signals propagate along the wire at a speed very similar to that of the soil or rock in the area if the wire is on or

near the ground. If the wire is elevated in the air then signals will travel at the speed of light (0.3m/ns).

Figure 9-24: Two examples of X events which travel along a wire near a transmitter-receiver GPR configuration. The strong

steeply dipping events show a partial X character. Signal attenuation is sufficiently high that the arms of the X don’t cross in these

examples.

This unique response can be seen when a metallic measuring tape is placed along the survey line and the radar system

is moved along beside the rape. Quite often people will use fabric or plastic tape measures which should not show

this effect. Unfortunately, some manufacturers embed thin metallic strength strands into fabric tapes. GPR users

should examine measuring tapes closely to avoid this potential problem. The examples in Figure 9-24 show X events

caused by 50 m tape measure with a hidden metal strength strand.

If you see this type of GPR response be aware of the mechanism that causes it and make sure that the tape measure

that you are using is not the source of the problem. If this is not a factor then look around you for a buried wire or

cable or a broken down fence which may be the source of the problem. Keeping antennas as far as possible from the

cable or wires will minimize impact.

Figure 9-25: Events on a GPR record are caused by signals which couple into a nearby wire or cable. Signal A travels from the

transmitter along the wire to the left end, reflects and travels back to the receiver. Signal B travels from the transmitter along the

wire to the right, reflects from the end, and travels back to the receiver.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



190

9.6 Plastic Water Pipe Diameter Determination

GPR users often ask whether GPR can determine the diameter of a pipe which has been detected. In many situations,

the pipe diameter is smaller than the GPR resolution length. As a result, estimating diameter is difficult.

Specific applications occur where the pipe diameter can be determined even where the diameter is small compared to

the resolution length. Water filled plastic pipes are a prime example.

The GPR response of a non-metallic pipe is dictated by the material filling the pipe, not the pipe composition. For

water filled non-metallic pipes, the water controls the GPR response. Water has a very slow GPR velocity causing

the signal transmitted through the pipe to be refracted diagonally across the pipe. The concept is shown in Figure 9-

26.

The slow velocity means that the travel time for the signals across the pipe is quite long. Reflection from the opposite

side of the pipe will be substantially delayed.

Figure 9-26: A water filled pipe represents a unique GPR target. The low velocity causes the transmitted energy to be refracted

diagonally across the pipe.

Many GPR records show detection of the top and the bottom of a water filled plastic pipe. The difference in travel

time between the top and the bottom responses provides a way to estimate the pipe diameter.

The rays representing the reflections from the top and the bottom of the pipe are depicted in Figure 9-27. The reflec-

tion from the top and the bottom are locked in time with respect to one another by a fixed delay time.

8/20/2019 GPR Annan 2003



191

Figure 9-27: When one traverses a pipe, the GPR signal is reflected from the normal point on the pipe’s circumference. Transmit-

ted energy traverses diagonally to the opposite wall of the pipe and is reflected. Rays always pass through the pipe center as

sketched.

The GPR response will be as depicted in Figure 9-28. The top and bottom of the pie yield identical hyperbolas offset

in time. Where water is very fresh, multiple echoes of the signal traveling back and forth in the pipe may be detected

as shown in Figure 9-28.

Figure 9-28: Travel time for multiple reflections in a water filled pipe. The top hyperbola represents the reflection from the top of

the pipe whereas the three other hyperbolas occur at a fixed delay time (30 cm diameter, 0.5 m deep).

8/20/2019 GPR Annan 2003



192

The unique feature of the response is that the top and bottom response track one another with a fixed time delay

between them. This time delay is determined as

where h is depth to the top of the pipe, v is velocity in the ground, x is offset of GPR from pipe axis, t is the time for

signal to do a two way transit through the pipe and n is the number of two way transits.

where d is pipe diameter and vw is water velocity (0.033 m/ns or 0.11 ft/ns).

An example GPR response traversing perpendicular to a 6” diameter water filled pipe is shown in Figure 9-29. Look-

ing closely one can see a hint of a third hyperbola representing a second pass (first multiple) through the pipe.

Figure 9-29: GPR response over a 6 inch plastic water main. the hyperbolic response from the top of the pipe is followed by a second response from the bottom of the pipe. The delay time of about 8.2 ns is the delay time for a 6 inch water pipe.

Figure 9-30 shows a profile along the axis of a pipe where the top and the bottom reflection track each other. The

apparent depth variations are caused by the fact the profile does not track perfectly along over the pipe but wanders

off to the side and back.

v

hT

tnTv

x

Tn

2

4

0

21

2

02

2

=

∆+

+=

∆

2

2 tvdand

v

dt w

w

∆==∆

8/20/2019 GPR Annan 2003



193

Figure 9-30: This Noggin SmartCart record shows a traverse perpendicular to a pipe followed by a traverse along the pipe. The

top and bottom reflection of the pipe are clearly visible.

The following table provides a summary of travel time versus pipe diameter. A quick estimate of diameter isobtained by measuring delay time between the events on the GPR record.

9.7 Antenna Shielding

What is antenna shielding? Figure 9-31 shows the conceptual idea of a shield. With GPR, the antenna is normally

places close to the air-ground interface. The shield is a “container” which encloses the antenna. The objective is to

improve the antenna’s performance by placing the shield over it.

Figure 9-31: Concept of a GPR antenna shield. The shield encloses the antenna to minimize coupling with signals in the air.

Diameter (cm) Diameter (in) t (ns)

5 2 2.7

10 4 5.4

15 6 8.1

30 12 16.2

50 20 27

∆

8/20/2019 GPR Annan 2003



194

What is the purpose? Referring to Figure 9-32, signals can travel from a transmitter to a receiver along a number of

paths. Shielding is used to achieve the following goals:

• maximize the energy on the path AA’ to and from the subsurface target

• minimize the direct transmitter to receiver energy on path B

• minimize the energy which escapes into the air as on path CC’• minimize external electromagnetic noise as indicated by signals D

Figure 9-32: A GPR system emits and detects radio wave signals. There are many possible signals and paths. The objective is tomaximize the target response and minimize others.

Given these laudable benefits, what are the drawbacks? Antenna shielding requires a structure that has an electro-

magnetic response. If the structure is not designed properly or is damaged, its electromagnetic response can be large.

In addition to the signals shown in Figure 9-32, energy goes from the transmitting antenna to both the transmitter

shield and the receiving antenna shield and then to the receiving antenna as indicated in Figure 9-33. the shield gen-erated signals can be large and reverberate for along period of time.

Figure 9-33: Antenna shields must interact with radio waves to be effective. The shield can generate responses which may be det-

rimental and interfere with the desired measurement unless extreme care is taken in the shield design.

Besides the electromagnetic response of the shield itself, an effective shield has to be larger than the antenna. Thisleads to considerable transducer size, weight and manufacturing cost penalties. In a nutshell, there is no free lunchand attempting to shield an antenna can create as many problems as benefits.

A series of radar traces are shown in Figure 9-34. (a) is the ideal response without external noise sources. (b) showsthe response observed when external noise is present. (c) shows the behavior which can occur if improper shieldingis used. The reverberatory signal running down the trace for a long period of time is the transient response of theshielding. (d) shows the result that a practical shield can achieve.

8/20/2019 GPR Annan 2003



195

(a) Unshielded noise free response.

(b)Typical response with external noise.

(c) Transient response when shielding is improperly implemented.

(d) Results that can be expected with a practical well implemented shield.

Figure 9-34:

8/20/2019 GPR Annan 2003



196

Shielded antennas can be readily constructed for higher frequency GPR systems, typically in the 100 MHz frequencyrange and above. The shields are about the same size as the antenna and use absorbing material to damp out theundesired signals.

At lower frequencies, practical size and portability dictates minimal or no use of shielding.

Tips for GPR users are as follows:

a) Shielding is never perfect no matter what some may claim.

b) Question whether your application requires shielding. the highest fidelity and maxi-mum depth of penetration at open sites may be obtained when shields are not used.

c) Always look for GPR events attributable to spurious signals. Even with the mostideal shield, spurious signal leakage can and will occur.

9.8 Ringing on Radar Records

Ringing is a term generally used in the GPR community for signals which reverberate in a regular fashion. Suchresponses are created when a GPR signal interacts with an object in such a way that the signal repeatedly bouncesaround within an object or between two or more objects.

The most common acoustic analogy is a bell. When a bell is tapped with a hammer, it will ring for a considerableamount of tie. The elastic energy causes the bell to deform in a rhythmic fashion as it propagates around the bell.The sound, which has a fixed tone determined by the bell size and composition, then slowly dies out.

Reverberation or ringing represents a resonant condition in the system. The general concept is shown in Figure 9-35which illustrates the transient response of an object.

In Figure 9-35, the upper trace (a) shows the input signal which is an impulse and the lower trace (b) shows an oscil-lating signal which lasts for a considerable period of time.

(a)

(b)

Figure 9-35: Example of a reverberating response. (a) The input excitation shown as an impulse. (b) The target behavior oscil-

lates and decays slowly with time.

8/20/2019 GPR Annan 2003



197

Sources of ringing in GPR measurements are many. Some common ones are:

• an undampened dipole antenna (Figure 9-36(a))

• inappropriately constructed antenna shielding (Figure 9-36(b))

• a metal object in close proximity of the antennas (Figure 9-36(c))

• a strong target near the antenna or ground interface (Figure 9-36(d))

(a) Undampened electric dipole antenna will support electrical current which travels back and forth repeatedly gener-ating signal.

(b) Currents generated on antenna shield bounce back and forth emitting signal.

(c) Current bouncing back and forth on metal object of similar size to antennas. See Figure 9-37 (a), (c) and (d).

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



198

(d) Partially transparent layer on structure which contains signal which can bounce back and forth. See Figure 9-37(e).

(e) Strong reflector generates signals which bounce back and forth between the object and the antenna (and/or shield,and/or ground surface). See Figure 9-37 (b).

Figure 9-36:

The fundamental period or repetition rate for the oscillation is a characteristic feature of ringing/reverberation. The

repetition rate is directly related to a spatial dimension of the object(s) causing the reverberation. The reverberation

period,

is a function of an object dimension, L, and also the velocity, v, in the object surrounding material. It is possible to

have a multiplicity of different dimensions yielding a multiplicity of periods causing a more complicated signal.

Generally one period will be dominant.

GPR reverberation is very selective because the excitation signal, object geometry and material properties must be in

accord to get a resonant response. Generally, the ground severely dampens out reverberation because it absorbs

energy, but it is quite common to see examples in GPR records. Figure 9-37 shows example of real survey data which

help provide insight as to what should be expected. The individual example captions indicate the source of the ring-

ing.

In addition to these examples in Figure 9-37, an example is presented in Section 9.7 Antenna Shielding.

v

Lt =

8/20/2019 GPR Annan 2003



199

(a) A recent 250 MHz survey exhibited a ringing event. A scrap piece of wire (above right) was found in grass)

(b) Example of multiple reflections on an ice road. The GPR signal bounces up and down several times because theice-water boundary is a very strong reflector and ice is transparent.

(c) A large buried metal object generates a low frequency (compared to GPR pulse) reverberating signal.

8/20/2019 GPR Annan 2003



200

(d) Two metal objects side by side with low and high frequency ringing.

(e) A target creates multiple equi-spaced responses that extend over a long time interval.

Figure 9-37:

9.9 GPR Time Zero & Depth Estimates

Interpretation of depths from Ground Penetrating Radar requires data knowledge of zero time in a GPR recording.

While this sounds like a trivial matter, precise location of zero time requires clear terminology and an understanding

of instrument operation.

Most important is the recognition that we are measuring signals that travel at the speed of light. We have to account

for the finite time that signals take when traveling through our measurement apparatus.

First, two definitions:

T0 = time zero - the time when the GPR transmitter initiates signal emission via the GPR transmitting antenna.

R 0 = first break - the time when the GPR receiving antenna detects the on-set of the signal traveling directly from

the transmitter.

8/20/2019 GPR Annan 2003



201

These two quantities are very important in GPR because R 0 is normally detectable on GPR records whereas T0 is not

recorded. The event determining R 0 normally travels on the direct path from the transmitter to receiver antenna at the

speed of light.

S - antenna separation c - speed of light (0.3 m/ns)

The key concepts are illustrated in the block diagram in Figure 9-38 and the sketch of a received signal in Figure 9-

39.

Figure 9-38: Simplified GPR system and measurement configuration illustrating signal time delays.

Figure 9-39: Illustration of a GPR received signal displayed as amplitude versus time.

c

STR += 00

8/20/2019 GPR Annan 2003



202

While we may have a very precise clock, the cabling and electronic devices in the transmitter and receiver have asso-

ciated time delays, t1and t2, which can be large. Unless every cable and circuit component time delay is mea-

sured, knowing when the transmitter emits and the receiver records is not possible.In normal GPR practice,calibrating cables and circuit delays is not cost effective. The procedure is to measure R 0 and infer T0 by knowing S.

When determining depth, the measurable times are R 0 and R 1. Quite often depth, d, is estimated by the simple calcu-

lation

(9-1)

while the correct solution is actually

(9-2)

which reduces to the first expression S=0 (the transmitter and receiver are spatially coincident). In the equation, v is

the propagation velocity in the ground.

When using GPR systems where antenna separation can be varied, operators should always record the value of S. For

fixed antenna geometries, accommodations for S can be designed into the instrument data recording stream. It is

always important to keep track of the value of S in reports.

For deep geological sounding, S << d and the approximate solution expressed in Equation 9-1 is perfectly satisfactory

for determining depth.

For non-destructive testing and buried utility applications, having is common and the exact solution in Equation

9-2 must be used.

9.10 GPR Antenna Elevation

GPR users invariably ask “Why can’t we lift the GPR off the ground?”, or “How close to the ground must the GPR

be?”

The simplest answer is that the best GPR data are normally obtained when the GPR antennas are closely coupled to

the ground.

So, the question arises: “Why does anyone ever want to raise the antennas off the ground?” The prime reason is that

some applications lend themselves to transporting the GPR system on a vehicle. In this case, the system can be

moved rapidly over the ground and large areas can be covered efficiently. Unfortunately, to move quickly over most

ground the antennas have to be raised so that the sensors do not get damaged when traversing a rough surface.

A compromise, based on an understanding of the trade-offs, must be reached. Lifting antennas above the ground has

operational benefits but is detrimental for many reason which are outlined below.

∆ ∆

( )2

01v

RRd

−=

( )( )( ) 2122

012

2

1 /Sc/SRRvd −−−=

S d≈

8/20/2019 GPR Annan 2003



203

What happens when the antennas are raised off the surface?

1. In many instances, antennas and electronics are designed to match the low impedance

ground and don’t always work as well when deployed in air.

2. When antennas are elevated, more energy goes into the air and less into the ground. The

antenna pattern variation with height is depicted in Figure 9-40 and Figure 9-41.

Figure 9-40: Illustration of antenna geometry over the ground. The critical parameter is the distance between the antenna ele-

ment and the ground surface, h.

Figure 9-41: Illustration of how antenna directivity changes with the normalized height of the antenna above the ground surface.

3. When mapping small objects, the response of small targets is substantially reduced as the

systems are elevated above the surface, as shown in Figure 9-42.

8/20/2019 GPR Annan 2003



204

Figure 9-42: Amplitude at the peak of a response decreases with elevation 10 cm diameter target with a 1000 MHz GPR system,

showing a substantially reduced sensitivity to a target for a GPR measurement.

4. When mapping small targets, the spatial resolution of the measurement is substantially

degraded with elevation of the GPR transducers.

5. When antennas are elevated more energy goes into the air and the measured results are

more prone to contamination by air wave events. This is depicted in Figure 9-43.

Figure 9-43: The data shown here were acquired with the GPR antennas elevated about one third of a wavelength above the

ground surface. The scattering from trees in the surrounding area (indicated by arrows) are quite visible on the record. Response from subsurface geology is within the area indicated by the thick line.

6. The GPR loses sensitivity for detecting steeply dipping targets such as geological hori-

zons, which can have substantial dip.

While there are many good reasons for keeping GPR antennas on the ground, some applications can tolerate antennas

being lifted.

http://-/?-

http://-/?-

8/20/2019 GPR Annan 2003



205

If the application has most of the following features then operating with antennas elevated off the surface may be per-

fectly satisfactory.

a) The targets being sought are large and relatively flat lying.

b) The survey area has few surface obstructions and has a flat surface.

c) The measurement environment has low loss conditions (so good penetration is

achieved).d) The targets are relatively shallow.

As a guide, good ground coupling occurs if the antennas are kept within 0.1 center frequency wavelengths of the sur-face, where the wavelength is in the material being probed.

In summary, conventional GPR gives best results when the transducers are kept close to the ground. Be sure to care-fully assess your application before operating with the antennas elevated off the surface as the results may not be asgood as you had anticipated.

9.11 Determining Layer Thickness & Velocity from CMP Data

In many GPR applications, CMP data acquisition is used to estimate the velocity in the subsurface. The most com-

mon method of analyzing CMP data is to use the classic “t2 - x2” method of extracting velocities and depths.

With modern computers, it is much more common for CMP data to be analyzed using a semblance approach whichgives a measure of coherence of GPR signal move-out versus antenna separation. Peaks in the semblance functionindicate the move-out velocity and arrival time for distinct reflections from the subsurface.

Translation of the semblance analysis data into true velocities and depths is sometimes required. The procedure for this is commonly known in petroleum seismic reflection. The equations involved are normally referred to as the Dixequation. This analysis was carried out by Dix (1956) and is a mainstay for determining interval velocities in the lay-ered structures.

9.11.1 Analysis Procedure

The basic concept of the layered earth giving rise to a number of reflections is depicted in Figure 9-44. The earth isassumed to be made of layers of arbitrary thickness and velocity.

Figure 9-44: In a simple layered earth environment, the variables are the layered thickness and the velocity of each layer. These

parameters need to be estimated from the GPR data.

8/20/2019 GPR Annan 2003



206

Figure 9-45: The layered sequence in Figure 9-44 would generate a CMP GPR section as shown above. The direct air and

ground waves are indicated as well as the hyperbolic reflection events from horizons 1, 2 and 3, which have intercept times of t 1 , t 2

and t 3.

When a CMP measurement is carried out in such a structure, one sees the idealized events arriving as depicted in Fig-

ure 9-45. Each of the horizons gives rise to a reflection event which closely follow hyperbolic move-out in the

antenna separation travel time space. Mathematically, the shape differs slightly from a hyperbola but petroleum seis-mic analysis in numerous text shows that this deviation is so small as to be inconsequential until extremely large

antenna separations occur.

The result is that an event from a flat lying horizon can be expressed mathematically as

(9-3)

where tn is the velocity with zero antenna separation and tn(x) is the travel time for a finite antenna separation x. The

value w is known as the move-out or “RMS” velocity.

w is a weighted mixture of all the velocities in the section above the horizon giving rise to the reflection. As a result,

it is not a true material velocity but rather a weighted sum.

Figure 9-46: This is an example from a CMP from an area where there is a strong water table reflection at a travel time of 70 ns.

The section above the water table has sands with varying water content well below saturation.

2

2

2

n

2

n w

xt)x(t +=

5 10 15 20 25 30

0

50

100

150

200

250

Separation (m)

T i m e ( n s )

5 10 15 20 25 30

0

50

100

150

200

250

Separation (m)

T i m e ( n s )

8/20/2019 GPR Annan 2003



207

An example CMP record is shown in Figure 9-46. This illustrates the raw radar traces with events occurring at vary-

ing times down the trace. Examining the record, it is clear that many of the events move-out in a hyperbolic shape

with time as antenna separation is increased. With modern computer analysis, a process called semblance analysis is

normally applied to CMP data which yields a measure called S(w,t). The semblance measure S is such that it will

yield a peak when there is an event in the GPR data which has a move-out velocity w.

Various forms of indicator can be used. Some of the common ones applied are

(9-4)

(9-5)

(9-6)

where

(9-7)

The r j in the above equations is the radar trace for antenna separation x j. The goal is to get a function S which peaks

sharply whenever there is a coherent event that moves out in a hyperbolic form in time-offset space.

∑

+=

j

j

j1 w

utr )t,w(S

∑ +

∑

+

=

j

j j

j

j

j

2

w

utr

w

utr

)t,w(S

+∑

∑

+

=

w

utr

w

utr

)t,w(S

j

j

2

j

j

j

j

3

2/1

2

2

j2

j w

xtu

+=

8/20/2019 GPR Annan 2003



208

Figure 9-47: This cross section is the semblance analysis for the data shown in Figure 9-46 . Peaks occur at travel times of 40 ns,

50 ns and 70 ns. RMS velocities are picked to be 0.95, 0.98 and 0.105(m/ns).

The normal analysis process is to generate a contour map of S as a function of w and t. An example of this is shown

in Figure 9-47. Data are interpreted by selecting the peaks in this function. The peaks identify particular reflecting

horizons. From the peaks in S one creates a table of values which are the zero offset time and RMS velocity for the

domain t reflecting horizons. One normally tabulates this as indicated in Table 1. The values tn and wn indicate the

intercept time and the RMS or move-out velocity for the horizon.

Table: 9-1 The Dix analysis for the layered sequence in Figure 9-47.

0.05 0.10 0.15 0.20 0.25 0.30

0

50

100

150

200

250

RMS Velocity (m/ns)

T i m e ( n s )

0.05 0.10 0.15 0.20 0.25 0.30

0

50

100

150

200

250

RMS Velocity (m/ns)

T i m e ( n s )

Dix Analysis for GPR Data

Horizon Intercept Stacking Interval Dix Interval

Number Time Velocity Thickness Depth Velocity

(ns) (m/ns) (m) (m) (m/ns)

0 0 0.000 0.00

1 40 0.095 1.90 1.90 0.095

2 50 0.098 0.55 2.45 0.109

3 80 0.105 1.74 4.18 0.116

Note: If the intercept and velocity data are not consistent, a negative interval velocity can occur and is

indicated by computation error #NUM!!!

Note: GPR velocities lower than 0.03 and higher than 0.3 are not realizable and indicate poor input

information.

8/20/2019 GPR Annan 2003



209

9.11.2 Intercept Time & RMS Velocity Conversion to Depth & Interval Velocity

For many applications it is satisfactory to estimate the depth of horizon by just using the simple equation

(9-8)

This gives a quick number which is quite often all that is required.

In some instances, however, obtaining a more correct velocity and depth model is desirable. This is the problem

which Dix addressed for reflection seismic many years ago. The approach can be viewed as peeling the onion layer

by layer; starting from the top, tn and wn are used to extract depth and velocity working down through the section.

The mathematical details can be found in many text. One good reference is Grant & West (1965) (pages 143-145).

To a very good approximation, the true velocity in layer n can be expressed in terms of the RMS velocity and inter-

cept times as

(9-9)

Once the velocity in layer n is known, one can determine the thickness of the layer. Mathematically this is expressed

as

(9-10)

Finally, the depth of each layer or horizon can be expressed as

(9-11)

Such calculations are very quickly implemented in spreadsheet analysis and can be integrated into normal data analy-

sis.

The layer velocity derived in this manner is called the “interval velocity”. This means this is the velocity in the inter-

val between depth n-1 and depth n. For some applications knowing the velocity in a particular layer enables estima-

tion of other parameters such as porosity, water content, density, etc. Again, extraction of this information requires

some prior knowledge of the geology.

For the data shown in Figure 9-46 and the semblance result in Figure 9-47, intercept times and RMS velocity for two

events have been tabulated. The resulting spreadsheet calculations are shown in Table 9-1.

In general, these calculations are easily carried out but there is considerable sensitivity to errors as larger depths are

2

wtd nn

n =

21

1nn

2

1n1n

2

nn2

n tt

wtwtv

−−

=−

−−

)tt(v

2

1h

1nnnn −−=

∑=φ=

n

inn hd

8/20/2019 GPR Annan 2003



210

explored. One can see from the nature of the velocity function that slight errors can cause gross variations in interval

velocities. Handling these errors is the trick of stabilizing the solution and requires constraints which are often based

on understanding the physical setting from where the data are acquired with GPR velocities less than 0.033 m/ns are

highly unlikely and greater than 0.3 m/ns impossible. Care should be exercised when using the Dix analysis outputs.

9.11.3 Direct Air Wave & Ground Wave

Unlike seismic applications, GPR quite often has two other events on the CMP record which are strong signals,

namely, the direct air wave and the direct ground wave. Both these waves run parallel to the air-ground interface and

arrive with zero offset times of zero. When carrying out semblance analysis S(w,t) will often show maxima at (v1,0)

and (c,0) where c is the speed of light. Sometimes the semblance processing will mute these events out.

This information is not used in the Dix analysis, but the direct ground wave event which yields at maximum at (v 1,0)

can corroborate the velocity determined for the first layer of the stack.

Often a thin veneer of higher velocity material may be present near surface and the reflection from this horizon

merges with the direct ground wave such that it is not resolvable in the t2 - x2 analysis. In this case, the velocity value

obtained for the direct ground wave may be more representative at very shallow depths than that which is derived

from the first reflection interval velocity which is averaging a thicker zone.

9.11.4 Summary

The preceding has been a quick overview of transforming CMP data into velocity versus depth. The Dix style analy-

sis and seismic is directly applicable to GPR as these above results show.

Using this type of analysis the interpreter should be cautious about over interpreting interval velocities and thick-

nesses as errors can arise in unpeeling the layered earth stack.

8/20/2019 GPR Annan 2003



211

9.12 CMP Analysis

VELOCITY MEASUREMENT

Ground penetrating radar (GPR) measures the time for signals to transmit to and from a target. The target depth or

distance is determined by multiplying travel time by velocity in the host material, which is frequently unknown.Common mid-point (CMP) soundings are used to measure velocity and hence translate regular GPR time observa-

tions into reliable depth estimates.

The below illustration show how a CMP measurement is made and the principles of how radar signal arrivals are used

to estimate velocity. Signals can follow a variety of paths from the transmitter (Tx) to the receiver (Rx). In a CMP

sounding the antenna geometry is varied in a controlled fashion so that signal arrival times follow predictable mathe-

matical forms enabling velocity estimation.

In the below example, the arrival time of Aw with distance confirms the speed of light, c, to be 0.3 m/ns while Gw and

R 1 indicate ground velocity, v, is 0.1 m/ns.

8/20/2019 GPR Annan 2003



212

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 10-Case Studies

213

10 CASE STUDIES

10.1 MINING & QUARRYING

ANISOTROPY IN COAL

Interest in the use of ground penetrating radar (GPR) for coal mining is widespread. Coal is frequently transparent to

radio wave signals and GPR enables mine engineers to quickly and easily locate and avoid potential problems, such

as pockets of methane gas, preventing accidents. GPR is also used to map unstable roof conditions.

While coal can be extremely transparent to GPR signals, it is important to realize that coal often has a grain much like

wood, and only produces results when surveys are conducted "against the grain". This study demonstrates the

extreme effects of antenna orientation on GPR measurements in coal.

The two data sets below were acquired using a pulseEKKO GPR system with 450 MHz antennas, with two antenna

orientations. The first data set was collected with the antennas running perpendicular to the coal seam (against the

coal grain). This approach proves successful as the data show a strong return from an adjacent entry. The second sur-

vey was conducted with the antennas parallel to the coal seam. With this antenna orientation GPR does not detect theentry or define any features in the coal.

These highly varied results from using different antenna orientations are a valuable lesson in effective use of GPR in

coal mining applications. This study demonstrates the effectiveness of GPR for coal mapping and indicates how to

prevent time loss and undesirable results due to using the improper survey procedure.

Data compliments of United States Bureau of Mining, Pittsburgh, USA Fracture Mapping

8/20/2019 GPR Annan 2003


10-Case Studies GPR Principles, Procedures & Applications

214

LIMESTONE CAVITIES

Ground penetrating radar (GPR) is frequently used to detect cavities in rock for construction planning. Limestones

are transparent to GPR signals and the karstic cavities (filled with air or water) common in limestone give strong

radar return signals.

At a building site in Italy, located on cavity prone limestone, the geotechnical planning necessitated locating cavity

formations before building construction could begin. The data below were acquired using a pulseEKKO GPR system

with 100 MHz antennas. The data clearly indicates open cavities one to two meters in diameter located throughout the

limestone to a maximum detection range of eight meters.

Following the survey, excavation work confirmed the radar interpretations. This information enabled the site engineer

to adapt construction accordingly, avoiding mishaps due to building on weakened limestone.

Data compliments of IGS, Italy.

8/20/2019 GPR Annan 2003



215

GRANITE QUARRY MAPPING

Efficient rock quarry operations require extensive knowledge of the rock conditions within a site.

Granite quarry sites are frequently mapped using ground penetrating radar (GPR) systems. GPR detects bedding

planes, fractures and variations in rock type which help rock mechanic engineers plan and expedite extraction.

A pulseEKKO GPR system with 100 MHz antennas was used in this case to quickly and efficiently obtain data. The

data set shown below clearly identifies a major fracture in the rock to a depth over 12 meters. Minor cross jointing is

also visible in this section. These data were acquired in winter when the surface was covered by two meters of snow.

Following the survey the granite was successfully extracted using information from the GPR data collected. The data

clearly identified optimal locations to mine, making granite block extraction more dependable and predictable.

Data compliments of multiVIEW Geoservices Inc.

8/20/2019 GPR Annan 2003



216

POTASH MINE SAFETY

Ground penetrating radar (GPR) is a powerful tool for mapping mine geology. GPR can define geology in all direc-

tions around underground structures and help assess rock mass stability. The study demonstrates the efficiency of

GPR for this application.

In potash mines in Saskatchewan, Canada, mining engineers regularly need to measure the thickness of salt over the

mine entries and whether or not water is present in the salt formations. If major geological deformations occur, min-

ing is not viable. A pulseEKKO GPR system with 100 MHz antennas was used by the crew to survey the mine roof

using a mechanical arm to hold the antennas on the back. The data show that the geology in this section of the mine is

fairly flat-lying, indicating the area is stable for mining activity.

GPR saves development time and ensures that mining areas are safe and viable.

Data compliments of Potash Corporation of Saskatchewan, Canada

8/20/2019 GPR Annan 2003



217

SALT MINE GEOLOGY

Salt mines are usually located where geologic strata are flat lying or gently sloping. Mining operations are often not

viable if major geological deformations occur so mine production engineers require rapid methods to map the geolog-

ical structure ahead of mining.

Ground penetrating radar (GPR) systems have been used for this purpose in salt and potash mines globally. The

below data are from a mine located in southwestern Ontario, Canada. Salt is mined at a depth of 1760 feet below

ground. Normal practice involves removing 48 feet (18 m) of salt below the mining level.

The section shown was acquired in less than 15 minutes and the ground conditions could be monitored by mine per-

sonnel throughout the surveying operation. The data were collected using a pulseEKKO 100 system with 100 MHz

antennas.

100 MHz pulseEKKO radar survey data along an entry floor in a salt mine in Southwestern Ontario, Canada. The

boundary of interest is the contact between the salt and a water bearing dolomite at a depth of 62 feet (19 m). The

continuity of reflectors within the formation indicate that the area will be suitable for mining.

8/20/2019 GPR Annan 2003



218

10.2 GEOTECHNICAL & ENVIRONMENTAL

FRACTURE MAPPING

Ground penetrating radar (GPR) is regularly used to survey potential underground nuclear waste storage sites. Gov-

ernment and nuclear agencies worldwide seek to determine the feasibility of storing hazardous nuclear bi-products in bedrock.

This study was conducted at an underground research laboratory operated by Atomic Energy of Canada. The survey

was aimed at identifying suitable locations for storing the nuclear waste at this site, without fear of extensive leaching

through fractures should a leak occur.

The below data were obtained using a pulseEKKO GPR system with a 50 MHz antenna along a 25 meter survey line

in an underground entry.

Following the GPR survey, it was determined that the site was not viable for nuclear waste storage. The GPR survey

located water bearing fractures in the granite, making the site unsuitable as leaching could occur.

8/20/2019 GPR Annan 2003



219

MINE SHAFT DETECTION

Ground penetrating radar (GPR) is often used to determine whether or not shallow mining activity has taken place.

This study was conducted near Nottingham, England, in a three acre parcel of land with a known coal seam located

subsurface. The objective was to determine if any mining activity had occurred in the coal seam.

A pulseEKKO GPR system was used with a 50 MHz antenna to survey this area. The system quickly and easily col-

lected the below section.

Although surface topography and mining records show no indication that the coal has been worked, the data indicates

otherwise. The major disturbance at 21 meters is generated by an old shaft leading down to the coal seam.

The subsurface disturbance determined by the GPR survey defines hazardous ground below the entry, and localizes

the site of imminent collapse.

Data compliments of Structure Testing Services (UK) Ltd.

8/20/2019 GPR Annan 2003



220

GEOLOGICAL STRATIGRAPHY

A major application of ground penetrating radar (GPR) is mapping geologic depositional history. This new advance-

ment enables geologists to relate current stratigraphy formation to that of older deposits. This study was conducted to

determine the effectiveness of GPR in defining depositional history.

A pulseEKKO GPR system with 50 MHz antennas was used to quickly and accurately survey the William River

Delta on the southern shores of Lake Athabaska, Saskatchewan, Canada. The data clearly detect and display numer-

ous depositional layers, erosional inconformities, and cross bedding.

This important application of GPR provides valuable insight to geologists studying groundwater flow, contaminant

transport, as well as petroleum engineers dealing with similar geologic conditions in reservoir rocks at greater depths.

Data compliments of University of Calgary, Department of Geography.

8/20/2019 GPR Annan 2003



221

LIMESTONE CAVITIES

Ground penetrating radar (GPR) is frequently used to detect cavities in rock for construction planning. Limestones

are transparent to GPR signals and the karstic cavities (filled with air or water) common in limestone give strong

radar return signals.

At a building site in Italy, located on cavity prone limestone, the geotechnical planning necessitated locating cavity

formations before building construction could begin. The data below were acquired using a pulseEKKO GPR system

with 100 MHz antennas. The data clearly indicates open cavities one to two meters in diameter located throughout the

limestone to a maximum detection range of eight meters.

Following the survey, excavation work confirmed the radar interpretations. This information enabled the site engineer

to adapt construction accordingly, avoiding mishaps due to building on weakened limestone.

Data compliments of IGS, Italy.

8/20/2019 GPR Annan 2003



222

GEOLOGIC HAZARDS

Sink hole development can be hazardous particularly in residential areas. Sink holes commonly form in areas under-

lain by limestone containing cavities; such limestone is called karstic limestone. The cavities form when groundwater

flow dissolves and washes away parts of the limestone and can grow until they reach the surface causing sink holes.

Ground penetrating radar (GPR) provides a quick and easy means of imaging karst features. GPR data often aid in

construction and rehabilitation of roads and buildings by providing advanced warning of hazardous situations.

This GPR survey was conducted in a residential area where a surface collapse had recently occurred. One resident

was concerned that his home was in danger, and wanted to know if a collapse was imminent near his house. A pul-

seEKKO GPR system with 100 MHz antennas was used to quickly and accurately map subsurface conditions. The

survey indicated that there was no problem near the house but located a cavity about five meters below the surface in

the backyard.

8/20/2019 GPR Annan 2003



223

OVERBURDEN STRATIGRAPHY

Ground penetrating radar (GPR) systems are used worldwide to delineate soil stratigraphy and determine depth to

bedrock for engineering purposes. Applications include road, rail and pipeline design as well as engineering planning

for expansion of dam sites and buildings.

This study was conducted in an area of glacial outwash sand and gravel deposits near Marathon, Ontario, Canada.

The survey objective was to determine the overburden stratigraphy at the site of a pulp and paper mill slated for

expansion. A pulseEKKO GPR system with 100 MHz antennas was used. The system was operated in winter

weather conditions by a two person crew.

The information obtained from the GPR survey facilitated locating and construction designs for the pulp and paper

mill waste water storage area.


8/20/2019 GPR Annan 2003



224

DETECTION OF BURIED TANKS

Ground penetrating radar (GPR) systems are often used to find buried tanks. Environmental concerns and avoiding

construction surprises necessitate locating USTs (underground storage tanks) before engineering site work com-

mences.

A GPR system was used here to find the exact position of tanks, without disturbing the soil above. GPR provides a

safe, quick and non-destructive means to find buried materials without putting crews at risk of contamination.

The data were collected using a pulseEKKO system with 200 MHz antennas.

Data compliments of A-Cubed Inc.

8/20/2019 GPR Annan 2003



225

UNDERWATER MAPPING

Ground penetrating radar (GPR) technology can easily penetrate freshwater and map objects below the surface. This

new application of GPR technology can reduce the use of conventional underwater mapping methods, which are inef-

ficient in shallow water.

This experiment was conducted at Cornell university to test the efficiency of GPR in mapping objects underwater. A

pulseEKKO 1000 GPR system with 450 MHz antennas was used to locate two small plates suspended at 150 and 190

cm below the water surface. The plates were suspended from a raft at different angles, and the survey was performed

on the raft. The GPR system successfully and accurately located the 6 mm thick plates, and also located the river bot-

tom at the correct depth.

The success of this experiment proves that GPR is not only a viable means of mapping underwater objects, but is also

faster and more reliable than conventional methods for shallow freshwater lakes and rivers. This new underwater

capability expands the applications of GPR to include pipeline locating, police investigations, environmental investi-

gations, and treasure mapping.

Data compliments of Cornell University, Department of Agriculture & Biological Engineering, USA

8/20/2019 GPR Annan 2003



226

WATER TABLE MAPPING

Ground penetrating radar (GPR) can efficiently map groundwater conditions and benefits for groundwater and con-

taminant hydrogeology studies.

This survey was conducted near Lake Superior in Ontario, Canada, when planning an expansion of a pulp and paper

mill. A pulseEKKO GPR system was used with 100 MHz antennas to accurately map the groundwater conditions and

the depth to bedrock.

The information obtained from the survey allowed the engineers to select optimal sites for the waste water lagoons

minimizing the risk of toxic chemicals leaking into surrounding groundwater.


100 MHz radar data mapping water table and bedrock is glacial sand and gravel deposits.

8/20/2019 GPR Annan 2003



227

CONTAMINATED GROUNDWATER

Ground penetrating radar (GPR) provides a powerful means of monitoring the remediation of contaminated ground-

water caused by salts or leachates, hydrocarbons or other organic materials.

This pulseEKKO GPR survey was conducted near a municipal landfill site in Ontario, Canada. Contaminants leach-

ing from the landfill are transported by groundwater flow to a nearby stream. Knowing the lateral extent and depth of

contamination ensures optimal installation of monitoring and remediation wells.

A pulseEKKO GPR system using 50 MHz antennas determined the extent of contamination, which is indicated by

the absence of signals on the first data set above.

Repeated at time intervals, GPR surveys allow the impact of remediation measures to be monitored. As shown in the

lower data set, this site remediation was quite successful. After five years some areas initially opaque to GPR signals

have become transparent.

8/20/2019 GPR Annan 2003



228

BATHYMETRY & SUB-BOTTOM PROFILING

Ground penetrating radar (GPR) is frequently used for mapping water depth and sub-bottom stratigraphy in freshwa-

ter lakes and rivers. Surveys are usually conducted from boats or rafts. In areas where open water freezes, surveys

can be carried out through the ice.

This study was performed on a lake in Northern Saskatchewan in conjunction with site surveys for uranium mine

development. The measurements were made using a pulseEKKO GPR system with 50 MHz antennas. The pul-

seEKKO GPR system was mounted on a sled and towed by a skidoo which facilitated rapid reconnaissance work.

The above section shows the soft sediment to rock boundary. This information was used to locate optimal areas to

deposit mine tailings.

Data compliments of Cigar Lake Mining Corporation & Golder Associates Ltd., Canada

8/20/2019 GPR Annan 2003



229

THE NOGGIN GOES SWIMMING

A Noggin 250 recently undertook a waterborne survey. Although the mode of operation is not what we normally rec-

ommend for our Noggin products, the results were quite impressive. Noggin systems are designed to be very water

resistant but they are not truly designed for submersion.

At a site in Oregon, Brian Herridge of 3Dgeophysics.com and Scott Mills from GeoDesign of Portland, Oregon, were

confident GPR could image bottom stratigraphy in the Umqua River.

Information about the bottom stratigraphy was necessary as part of a geotechnical design project to be undertaken by

Brown and Caldwell and GeoDesign. Using GPR provided a more environmentally friendly and lower cost solution

than drilling.

The Noggin was placed directly in the water and towed around using a small paddle boat (see picture below). Several

lines about 100 feet in length were surveyed with quite spectacular results. An example profile is shown below.

These raw Noggin PCX images show water depth and subsurface stratigraphy to a depth of about 9 feet (3 metres).

These results are typical of what can be obtained in freshwater conditions; water is transparent to GPR systems when

the total dissolved solids (TDS) are low. As well, the slow speed of radio waves in water gives superb depth resolu-

tion.

There are many geotechnical problems which call for imaging water depth and sub-bottom stratigraphy in shallow

lakes and rivers. Normally we would recommend putting the Noggin into a small floating boat made of fibreglass,

plastic, wood or rubber. The Noggin can sound right through the bottom of the boat structure and see through the

water into the sub-bottom. This makes for a very efficient way of carrying out shallow water surveys. Noggins can

be used in this way in any number of applications. The Noggin is so simple to deploy, you only need to let your

imagination define a potential application and then try it out.

8/20/2019 GPR Annan 2003



230

Warning: Noggins will not see through sea (salt) water. The energy from the system is rapidly absorbed and very lit-

tle penetration will be achieved. Only use Noggins in freshwater with low TDS counts. Contact our professional

staff if you have any questions about Noggins and their use in water.

8/20/2019 GPR Annan 2003



231

10.3 FORENSIC & ARCHEOLOGY

OLD CATHEDRAL FOUNDATIONS

Archeologists use ground penetrating radar (GPR) to map dig sites because radar anomalies accurately pinpoint

hotspots to be investigated. Area surveys provide a powerful means of detecting the extent and shape of buried struc-tures.

This GPR study was conducted in Denmark, where an old cathedral foundation was known to exist underneath the

modern city. Parts of the cathedral had been discovered when waterlines were installed, though the extent and loca-

tion of the foundation remained unknown. A pulseEKKO GPR system with 200 MHz antennas was used to effi-

ciently map the extent and location of the foundation. The data collected clearly show the foundations and also locate

the waterline.

The resulting survey information facilitated future city planning to avoid disrupting the historical structure.

8/20/2019 GPR Annan 2003



232

SUBSURFACE STRUCTURE MAPPING

Ground penetrating radar (GPR) provides a powerful means of mapping subsurface structures. Where many existing

maps are old and inaccurate, GPR can provide a precise account of utilities, tunnels, pipes and cables, as well as sur-

rounding features. This is important information to have for future planning and development.

The below 50 MHz pulseEKKO results from Sweden accurately locate the tunnels cut in bedrock at 11 meters below

surface. A cable and a culvert at one meter depth are also detected. The GPR information indicates minor fracturing

in the surrounding bedrock, which is invaluable information for detecting potentially hazardous situations.

The efficiency of GPR subsurface structure mapping is a significant advancement for city planners, civil engineers

and developers. It not only informs of existing structures underground, but also warns of potentially hazardous situa-

tions enabling proper precautions to be taken before any unwanted situations arise.

Data compliments of TS Geokonsult (Sweden).

Ground penetrating radar clearly locates the two tunnels shown in the diagram, represented by the two arches on the

left side of the data set. The two shallower features on the right side are generated by a cable and culvert approxi-

mately 1 m deep.

8/20/2019 GPR Annan 2003



233

GRAVE LOCATION

Location of unmarked graves in cemeteries is an unusual but regular application of ground penetrating radar. Where

there are no markers or indications of grave locations development, construction projects can be quickly brought to a

halt until the history of the site is evaluated.

Ground penetrating radar (GPR) systems respond to buried objects as well as disturbed soil, providing a powerful

method to define unmarked grave locations. Other applications include forensic investigations and archaeological site

evaluations.

In this case a pulseEKKO GPR system was used with a 200 MHz antenna. The data were acquired by students at the

University of Calgary. The portable nature of the pulseEKKO system made data acquisition quick and easy. The soil

at the site was silty sand.

Data compliments of University of Calgary, Department of Geography, Canada

8/20/2019 GPR Annan 2003



234

THE LOST SQUADRON

Ground penetrating radar (GPR) is commonly used in glaciology, as ice is easily penetrated by GPR signals.

GPR enabled the recovery of The Lost squadron in Greenland. A squadron of aircraft were forced to land in Green-

land during WWII, although the exact location remained a mystery. The Lost squadron, consisting of six P-38F

Lightning fighting bombers and 2 B-17E "Flying Fortresses" were last seen on the Greenland ice in July 1942.

A one-man survey crew used a pulseEKKO system to map out the exact location of the squadron in May 1992, 50

years after the squadron's disappearance. The pulseEKKO system was excellent for use in this case because of the

lightweight, modular design enabling one person to use it easily.

Following the GPR survey, the site was excavated and a nearly complete aircraft has been successfully recovered.

Data compliments of Greenland Expedition Society

8/20/2019 GPR Annan 2003



235

BURIED VIKING ROADS

Ground penetrating radar (GPR) is in use worldwide to map objects for archeological digs. GPR systems determine

the best locations to dig and map the layout of obliterated roads and buildings.

This study was conducted in Denmark, beside the Alling River et sjellbro. A pulseEKKO GPR system successfully

located and mapped an ancient Viking road which sunk into a peat bog. In 1952 a stone carving was discovered in the

meadow nearby, which marked the crossing of the ancient road system. Parts of the road were excavated from 1952-

57 and in 1980.

The construction of the road consists of four structures. The two lower portions are comprised of small rocks and

twigs indicating the road was built in the early iron age. The two upper sections are wooden structures, one being

constructed in the late iron age, around 756 AD, and the other in the Viking age at about 1000 AD.

The GPR system was used to efficiently locate the ancient road system, without disturbing the structure.

8/20/2019 GPR Annan 2003



236

FORENSIC INVESTIGATIONS

Ground penetrating radar (GPR) is frequently used by police for forensic investigations. The ability to quickly locate

soil disturbances, voids, and trenching, as well as both metallic and nonmetallic objects makes GPR a valuable tool

for use in forensic studies.

In this case, the police had a search warrant and were on a seizure mission for evidence in a racketeering sting opera-

tion. A GPR survey was used to find money, firearms, and narcotics said to be buried at a private premises. One per-

son surveyed the front and back yard, the garage as well as some forest area behind the property, using a pulseEKKO

100 system equipped with an odometer wheel and 200 MHz antennas. The data were displayed on a computer screen

in real-time, enabling the police crew to immediately excavate and tag the significant finds as evidence.

The profile above collected in the front yard of the home shows trenching on the left side as well as a buried metallic

object on the right. The pulseEKKO survey allowed police to quickly collect evidence to strengthen their case.

Data compliments of multiVIEW Geoservices, Canada

8/20/2019 GPR Annan 2003



237

IN SEARCH OF CAVES AND GRAVES IN ISRAEL

Harry M. Jol has a Ph.D. from the University of Calgary where he wrote his thesis on Ground Penetrating Radar. He

also has a M.Sc and a B.Sc from Simon Fraser University in British Columbia. His interests include geomorphology

and GPR, particularly in coastal, deltaic, aeolian and geoarchaeological settings. Dr. Jol is currently a tenured Associ-

ate Professor in the Department of Geography and Anthropology at the University of Wisconsin-Eau Claire. The fol-

lowing is written by Dr. Harry Jol.

The Dead Sea Scrolls, which were accidentally discovered in a cave by a Bedouin boy in 1947, are one of the greatest

manuscript discoveries of the twentieth century. The Scrolls reveal the history of the Second Temple period - a time

of important developments in monotheistic religions.

The Qumran and surrounding region (from Golb, 1995).

Since 1947 the Qumran region, on the north-western shores of the Dead Sea, Israel, has been subject to countless

probes. GPR has been used at the site on several occasions in the pst with R. Eisenman being the only one to publish

any GPR interpretations.

Encouraged by these earlier results, a multi-site expedition used GPR during the summer of 2001. The expedition

was initiated with the hope of better understanding the Qumran site.

8/20/2019 GPR Annan 2003



238

Cave 4, the site of one of the Dead Sea Scrolls discoveries. Circled is the entrance to the cave.

The GPR portion of the project had two major objectives: 1) to identify if there were more caves in the marl cliffs thatcontained artefacts associated with Qumran; 2) to determine the positions of previously unknown or unmarked gravesin the cemetery east of Qumran.

The site was culturally important and needed to be treated with sensitivity and care. Geophysical methods, such asGPR, provided just the right approach - non-invasive and non-destructive.

Portable and digital, Sensors & Software’s Noggin plus 500 system and the pulseEKKO 100 GPR system were used tocollect data. The Noggin SmartCart system was ideal for rapid data collection over the graveyard area while the lowfrequency pulseEKKO system was used in the search for caves. The application of radar stratigraphy analysis on thecollected data aided in identifying the geometry of possible buried archaeological features.

The Qumran site has a thin, electrically resistive fan deposit with a more conductive marl unit below. The site condi-tions reduced GPR penetration to approximately 1 to 2 m but allowed enough penetration to map disturbances andvoids in the near subsurface.

The first objective, to identify if there were more caves along the marl cliffs, was achieved by running pulseEKKO100 GPR surveys along the cliff tops and along the cliff faces. Individuals with antennae were lowered with ropes

along the cliff faces and surveys were conducted along predetermined lines. Two sites were chosen for excavation based on GPR images that showed hyperbolic features between 0.5 and 1.0 m depth.

A grave with surface expressions (stones piled up in oval pattern) just east of Qumran Archaeological site.

Once excavated, the first site showed a localized patch of pebbles at a depth of 0.5 m that had interfingered with the

marl deposits. The second site resulted in the recovery of archaeological artefacts from the floor of a cave that had

collapsed.

8/20/2019 GPR Annan 2003



239

GPR profile showing a “hyperbolic” pattern along the cliff face. As the site was excavated, artefacts were located from an inter-

preted collapsed cave. The horizontal line is at approximately 0.5 m depth.

GPR profile showing a “V” shaped reflection pattern that is interpreted as a potential gravesite. There is no surficial expression

of the site being a grave. Total depth is approximately 1.5 m.

The second objective was to locate potential travesties at the Qumran cemetery where burial stones had been removed

due to vandalism, construction projects, or use in other burials. Since no complete map of the Qumran cemetery had

been constructed, GPR was utilized to map the subsurface with the hope of identifying unmarked burial sites.

Initially, an experiment was conducted on known grave sites so that GPR reflection patterns could be identified. Two

patterns emerged as burial signatures - a hyperbolic feature and/or a “V” shape.

An extensive GPR survey was conducted along the outer edges of the exposed cemetery as well as empty patches of

ground within the exposed cemetery. Interpretation of the GPR profiles, as they were collected, located over 100 potential graves that did not show surficial expressions.

GPR proved to be an effective method for locating and mapping potential caves and graves at the Qumran site with-

out disturbing the historically important surroundings. Based on the results and discoveries of this initial study, fur-

ther GPR surveys will be conducted at the site during the summer of 2002.

Support for the Multi-site Expedition to Qumran was provided by the John and Carol Merrill Foundation,

Frankel Fund, Biblical Archaeology Society, University of Wisconsin-Eau Claire and Sensors & Software Inc.

8/20/2019 GPR Annan 2003



240

10.4 BURIED UTILITIES

PIPES & BARRELS

The below data section demonstrates the use of ground penetrating radar (GPR) for pipe and barrel location. The pul-seEKKO GPR systems clearly locate both metallic and nonmetallic objects. The data also displays the ability of pul-

seEKKO GPR systems to locate small objects, such as the pipes.

The hyperbolic shape (inverted U) of the GPR response is diagnostic of localized targets. The top of the hyperbola

indicates feature location. The shape of the tails gives a measure of velocity and depth.

The example data shown here were acquired using a pulseEKKO 1000 system with 450 MHz antennas. The data

were recorded using 16 stacks at 0.1 meter step intervals. Note the difference in response for the two barrel orienta-

tions indicate that GPR can distinguish geometry.

Data compliments of University of Waterloo, Canada

8/20/2019 GPR Annan 2003



241

UTILITY DETECTION

The accurate location of buried utilities is imperative for contractors. Hitting a high voltage power line or water main

can be dangerous, disruptive to citizens, as well as expensive.

Ground penetrating radar (GPR) systems are used globally to locate metallic and nonmetallic utilities underground.

GPR's high resolution capabilities allow detection of closely spaced utilities which is often the case under roadways

in urban areas.

The below data were collected using a pulseEKKO GPR system with 450 MHz antenna on a major road in Kuala

Lumpur, Malaysia. The contractor needed to know where he could safely dig without hitting a utility while installing

fiber optic communication cables. The survey was carried out in a few minutes, and locations of the utilities were

marked on the pavement as the survey was being conducted. The fiber optic cable contractor quickly and safely

trenched between sites of buried utilities.

8/20/2019 GPR Annan 2003



242

TRACKING A STORM SEWER

At a recent Underground Focus Utility Workshop, we had the opportunity to see utility locating in action. The semi-

nars were held at Dekalb College in Covington, Georgia near Atlanta. The objectives of the workshop were to expose

utility locators to new technologies and practices in the industry and to provide hands-on experience with the various

techniques.

Figure 1: Location of storm drain intake.

Figure 2: Location of GPR lines.

After construction of the Dekalb College site, no as-builts were provided by the contractor. As a result, no records of

the underground utility pipes and cables were available. The technologies showcased at the seminar helped locate

some of the buried utilities.

Ground penetrating radar (GPR) proved to be very effective at the site. One interesting exercise involved locating a

storm water drainpipe under the circular driveway outside the lecture hall. While the intake could be seen at the curb,

8/20/2019 GPR Annan 2003



243

the location of the storm drainpipe was unknown. Figure 1 shows the circular driveway with the drain intake at the

edge of the curb.

Several GPR traverses around the intake located a large pipe. A series of GPR transects were then carried out to track

the alignment of the pipe. Three transects perpendicular to the pipe are indicated on the photo in Figure 2. The GPR

response on the three transects are shown as Lines 1, 2 and 3. The drainpipe was expected to continue down grade to

the right in the photo in Figure 1 and 2. Instead, the pipe ran in the opposite direction.

To address the drainage/slope issue, we then profiled along a transect (Line 4) which joined the pipe location deter-

mined from Lines 1, 2 and 3. This data provided a continuous profile along the axis of the pipe. The depth of the pipe

along the traverse is seen readily on Line 4 in Figure 3 and 4.

Figure 3: GPR survey lines over the storm sewer. Lines 1, 2 and 3 run perpendicular to the path of the sewer pipe. These lines

show the pipe at different depths, indicating that it is sloping. Line 4 runs parallel, along the length of the pipe, confirming the

slope.

8/20/2019 GPR Annan 2003



244

Figure 4: The GPR response from the storm sewer shows the reflections from the top and bottom of the pipe. Assuming the pipe is

air-filled, the pipe diameter was estimated at 36 inches.

From the transect in Figure 4, it was apparent that the depth of the pipe increased from right to left by about 3 ft (1 m).

In addition, it was possible to see the reflections from the top and the bottom of the pipe. The storm drain appeared to

be a concrete pipe with no apparent metallic structure associated with it. Using the depth of the top and bottom of the

pipe and assuming the pipe was air filled (there had been no rain for an extended period of time) it was possible to

estimate the pipe diameter at about 36 inches (90 mm).

This example illustrates the power of using GPR. The exercise of locating the pipe, marking the alignment and track-

ing the depth as well as estimating its diameter took about 10 minutes.

8/20/2019 GPR Annan 2003



245

10.5 STRUCTURE ASSESSMENT (NDT)

VOIDS BENEATH CONCRETE

Retaining wall, road and sewer collapses are commonly caused by soil erosion beneath or behind concrete structures.

Ground penetrating radar (GPR) can detect these unsupported void areas when they are in their infancy and help toeradicate hazardous collapses.

This experiment was conducted in College Station, Texas to test GPR for mapping voids beneath concrete. The Texas

Transportation Institute test site consists of an eight inch concrete slab with small voids from 1/8" to 1". A pul-

seEKKO 1000 GPR system with 900 MHz antennas accurately located all the voids quickly and easily, as clearly dis-

played in the data set below.

The success of this experiment not only confirms that GPR is an accurate means of void detection, but shows that

GPR can locate very small features, meaning that they can be detected and fixed before reaching a hazardous condi-

tion.

Strong reflectors are induced by relatively small air filled gaps under concrete.

8/20/2019 GPR Annan 2003



246

RAILROAD BED MAPPING

Safe and rapid rail service requires continuous maintenance of tracks and rail beds. Ground penetrating radar (GPR)

provides a quick and easy method of investigating subsurface conditions. The efficiency and speed of GPR surveys

permit remediation to start before conditions become hazardous.

Large sections of the trans-Siberian railway are constructed on a berm overlying permafrost terrain known to contain

ice wedges and lenses. If ground ice melts, rails can subside and possibly causing train derailment. A rail-mounted

pulseEKKO 100 GPR system with integrated wheel odometer attachment allowed for a quick and easy survey.

The survey determined the exact location of the ice wedge at 4243 meters along the survey line. The data also clearly

show the slumping due to the subsurface conditions and locates the ballast, railway embankment and peat layers.

The information obtained in this survey proved invaluable to the engineers as the location of the ice wedges was

determined before melting occurred. The situation was rectified before any damage was done to the railway equip-

ment or any passengers were put in danger.


8/20/2019 GPR Annan 2003



247

ROAD BEDS

Ground penetrating radar (GPR) is effective in defining potential problem areas on roadways. Data can be collected

quickly with vehicle mounted systems, and can accurately define subsurface conditions.

This study was conducted in Sweden to evaluate the cause of pavement failure. The pulseEKKO 1000 system was

used with a 900 MHz antenna to look through the roadbed and pavement to the bedrock below. The GPR data clearly

indicate cracks developing in the pavement which correlate with differential subgrade settlement. In this case, shal-

low bedrock is more stable than adjacent fill.

The information obtained from the GPR survey allowed for timely maintenance and repair of the roadway.

Data compliments of TS Geokonsult, Sweden

8/20/2019 GPR Annan 2003



248

REBAR MAPPING

Ground penetrating radar (GPR) is an invaluable tool for construction planning as well as road, pavement, and build-

ing maintenance and repair. The accuracy of GPR systems in the detection of rebar position and condition is impor-

tant in determining the stability of the structures. This study demonstrates the effectiveness of GPR for mapping

rebar.

This survey was conducted on the floor of a one story slab on grade building using a pulseEKKO 1000 system with

1200 MHz antennas. The data collected clearly show the rebar to a depth of eight to 15 centimeters, and show the

reinforcement curving up due to improper emplacement during construction.

The bottom of the concrete is weakly visible and is most readily detected when the rebar is closest to the surface.

8/20/2019 GPR Annan 2003



249

VOIDS BEHIND PIPES

Ground penetrating radar (GPR) is in use around the world for locating voids behind nonmetallic pipes and tunnels

because it can both find voids and estimate the size of the voids, even through reinforced pipe. Early detection of

voids is crucial to protect the structural integrity of the pipe and prevent extensive damage. The loss of material flow-

ing through the pipe is expensive and can pose an environmental threat.

This pulseEKKO survey was conducted in a one meter diameter reinforced concrete pipe to demonstrate the use of

GPR for void definition. A pulseEKKO 1000 GPR system with 1200 MHz antennas was used to define the pipe wall

as well as the surrounding areas. In this case the pipe walls consisted of 100 mm thick concrete with a double layer of

10 mm diameter rebar placed 50 mm apart. 50 mm and 25 mm thick pieces of 300 mm polystyrene squares were

placed outside the pipe to simulate voids.

The pulseEKKO data above clearly locate both the 50 mm and 25 mm thick polystyrene pieces and differentiates

between the thickness of each piece. The dots on the data between the travel time of 1.6 and 2.2 indicate the rebar

present within the concrete pipe.

8/20/2019 GPR Annan 2003



250

GPR MAPS FOUNDATION STRUCTURE

GPR can be used to map subsurface conditions with minimal impact on surrounding activity. The current surveyobjectives were to assess construction practices and to locate utilities in a hotel ballroom prior to renovation.

Figure 1: Photo of the doorway to the ballroom. Note the carpet on the floor is laid over concrete slab-on-grade.

To investigate, a Noggin 1000 Smart Handle system was deployed. The lines were laid out through a doorway toobtain “through-the-wall” data. The floor area was carpeted and underlying the carpet was concrete slab-on-grade.

Figure 2: The Noggin 1000 SmartHandle System configuration.

A photo of the doorway is shown above in Figure 1. This shows the position of the doors, the carpet over the floor and the survey line layout.

The Smart Handle system configuration is shown in Figure 2.

The grid acquisition mode for the Smart System enabled acquisition of data to image the subsurface in plan mapform.

A GPR cross section is shown in Figure 3. Note how this section through the doorway reveals the rebar as well as anembedded utility. A trench associated with construction of footings is also visible beneath the concrete.intervals. Theresulting data were processed using the EKKO_Mapper software to generate the plan map shown in Figure 4, whichslices through the subsurface between depths of 0 inches and 12 inches.

The depth slice clearly shows the rebar pattern and utilities location in the subsurface. Based on this, plans for reno-vation can be made with a better understanding of the prior construction practice.

8/20/2019 GPR Annan 2003



251

The above example clearly illustrates the power and the non-intrusive nature of GPR for building retrofit planning.

The data were acquired in 15 minutes while the ballroom was in use with no impact on daily activities. The initial plan map was created within a few minutes after data acquisition.

Figure 3: A sample GPR cross section through the middle of the ballroom doorway. Note the rebar, the utility lines and the trench

beneath the concrete.

Figure 4: GPR depth slice map between 1 and 12 inches, showing the plan layout of the conduit and reinforcing construction ele-

ments.

8/20/2019 GPR Annan 2003



252

CONQUEST GIVES THE CONCRETE PICTURE!

Conquest is specifically designed for non-destructive concrete inspection. Based on the Noggin 1000, Conquest is an

integrated GPR system which allows the user to gather data, process it and evaluate the results in the field. Down-

loading the information to a computer or laptop is an available (but not required) feature.

Figure 1: Conquest on the test sidewalk

There is increasing demand for mapping features within concrete before cutting or drilling occurs. Conquest can

locate conduit, rebar, and post-tension (PT) cables embedded in the concrete, as well as voids located beneath it. The

repair costs of cutting a conduit or PT cable can be high, not to mention the losses due to downtime.

The following example illustrates data collected from a 1200 x 1200 mm concrete test sidewalk located at Dofasco in

Hamilton, Ontario (Figure 1). Buried in this sidewalk were rebar and conduits of different sizes and composition(Figures 2 & 3). A plastic grid was overlain on the sidewalk and data were collected along the lines marked on the

grid.

8/20/2019 GPR Annan 2003



253

Figure 2: Items buried in the sidewalk

The scan was taken several months after the concrete had properly cured. Wet or freshly poured concrete is not

favourable for GPR measurements, due to the presence of free ions and water, which absorb the radar signal. Once

the concrete has cured, the ions become “locked” into the concrete structure, allowing increased penetration of the

radar waves.

After processing the raw data, Conquest generates plan maps at a series of depths in the concrete. The rebar and con-

duits are shown in the depth slice in Figure 4. One inch diameter conduits are seen at the bottom of the figure; two of

them are buried directly underneath rebar. The larger diagonal conduits are more visible in the deeper slice shown in

Figure 5.

By analyzing different depth slices, Conquest can help the user locate the position and depth of any embedded fea-tures. Drilling and cutting can then be carried out in the areas defined by the image that are clear of rebar and con-

duits.

Figure 3: Close-up of the scanned area. There are 5 conduits running from left to right all 1” diameter. From top to bottom, the

conduit materials are: galvanized steel, PVC, electro-metallic tubing (EMT), galvanized steel and PVC.

8/20/2019 GPR Annan 2003



254

Figure 4: Depth slice between 4.8” - 6.0”.

Figure 5: Depth slice between 6.0” - 7.2” The diagonal conduit on the left is 4” PVC, while the one on the right is 3” galvanized

steel.

8/20/2019 GPR Annan 2003



255

IMAGING OF A CONCRETE SLAB-ON-GRADE

Many problems such as assessing whether reinforcing is present or if conduits are embedded in the concrete require

examination of slab-on-grade concrete structures. Related problems include estimating slab thickness, detecting

voids beneath the concrete, and locating utility pipes, cables or tanks under the slab.

Figure 1: Photo of the test slab with axes indicated. A Noggin 1000 SmartCart maps the area.

The slab example presented here comes from the UK premises of Geomatrix Earth Sciences. A site photo is shown in

Figure 1. The 3.5 x 4.5 m slab thickness varies from about 140 to 180 mm (6 to 7 in) thick and has a steel mesh

embedded in it. The site has been examined any times with GPR but was only carefully studied recently. In Novem-

ber 2000 an accurately positioned GPR survey was carried out using a Noggin 1000 SmartCart system. Data was

acquired at 10 cm line spacings with 1 cm station intervals in both directions over the slab.

Figure 2: Depth slice centered at 100 mm (4 in) depth showing lapped sheets of reinforcing mesh. The mesh depth varies; out of

slice features appear as weaker responses. By looking at several depth slices, depth variation becomes more apparent.

8/20/2019 GPR Annan 2003



256

Figure 2 shows a 100 mm (4 in) thick depth slice centered at a 100 mm (4 in) depth. The image shows several over-lapping sheets of mesh are used to reinforce the concrete. The bright spot at coordinate (X = 3.5, Y = 1.4) is a voidunder the slab.

Figure 3 shows a similar slice centered at a depth of 500 mm (20 in) below the surface. In this slice a 150 mm *6 in)diameter plastic drainpipe is clearly visible.

The site flooded in the spring of 2001. This resulted in an increase in the water table levels in the local area as well asflooding over the slab and into the buildings. Subsequent to the flood, the site was resurveyed in early June 2001 todetermine if there had been any erosion of fill from under the slab. Figure 4 shows a depth slice of the fill material ata depth of about 300 mm. The distinct oblong feature was not observed previously.

The anomalous feature was drilled and was found to be caused by the presence of water saturated silt underlying theconcrete slab. Other areas of “ normal” response were underlain by coarse gravel fill.

The slab surveys have taken very little time - typically about 1.5 to 2 hours to survey once - and about 1 hour to pro-cess the data. The unique feature is that all equipment and processing is standard commercial product - not research-only technology.

Figure 3: Depth slice centered at 500 mm (20 in) which detects a 150 mm (6 in) diameter plastic drainpipe.

Figure 4: Depth slice at 300 mm (12 in) which shows the area after the site was flooded. An anomalous oblong area exhibits a

lack of signal.

8/20/2019 GPR Annan 2003



257

NOGGIN LOCATES SINKHOLE

Sinkholes are a problem in many geographic areas. Quite often we will see headlines reporting that a major structure

is suddenly sinking into the ground because a sinkhole is forming. The present subject of Noggin Notes describes the

mapping of a developing sinkhole in Florida.

The interesting case history is from a recent project of BCI Engineers and Scientists Inc. based in Lakeland, Florida.

BCI regularly conduct geotechnical investigations around residential and commercial structures. The current investi-

gation was carried out in October 2000 when BCI was contracted to investigate a depression in the ground surface

near a house.

BCI use a broad range of geotechnical tools for their customers. GPR is a standard part of their investigation arsenal.

The current work was carried out using a Noggin SmartCart with 250 MHz and 500 MHz transducers. the GPR sur-

vey consisted of 26 transects across the residential property; several transects ran directly across the shallow depres-

sion in the lawn. Figures 1 and 2 display some examples of these measurements. These images show the classic

GPR response of a sinkhole.

Both traverse images show a number of shallow point-like reflector responses (the inverted V’s on the record). These

responses are from utilities and irrigation pipes buried in the lawn.

The most widespread feature is the fairly continuous horizon indicated as B in the figures. This horizon is a sand-to-

clay boundary located at about 4 to 6 ft below the surface.

Figure 1: Example of a GPR traverse over a sinkhole: This image shows the depression of the sand/clay interface (B) with strong

localized (inverted U’s) responses (A) beneath the deepest point. Near surface utilities and irrigations pipes are visible as shallow

inverted U’s in the top 2ft.

The sand/clay boundary is ubiquitous and normally at a fairly constant depth. In the area where the depression wasnoted, the sand/clay boundary is pulled down and its response fades out in some locations. This depression of the

sand/clay interface into a funnel shape (if viewed in 3-D) is a strong indicator of subsidence at depth. Here the

boundary drops about 2 ft. deeper in the depressed area.

A second indicator of sinkhole development is the series of reflections at deeper depths that form beneath the depres-

sion in the sand/clay boundary. Such responses are usually localized in nature and indicate cavities which are gener-

ated from active revelling of the underlying soil. Example features are labelled A in Figure 1 and 2.

8/20/2019 GPR Annan 2003



258

Figure 2: A second example of traversing over a sinkhole. Note the absence of the sand/clay boundary reflection A at position

85ft. where deeper localized targets A are indicated. The subsidence blurs out the sand/clay interface which causes the GPR

reflection to fade.

The absence of a reflection from the sand/clay boundary is another indicator of sinkhole development. Lack of

reflection occurs in the area of most severe subsidence because the soil mixing in an area of active revelling creates

an indistinct interface.

This use of the Noggin SmartCart System demonstrates a very cost effective means of evaluating subsurface condi-

tions. The results are extremely graphic, the analysis can be made in the field, and the information readily inter-

preted by anyone with a little bit of experience using GPR in sinkhole terrain.

We thank all the folks at BCI for providing us with this very educational example.

8/20/2019 GPR Annan 2003



259

ON TRACK WITH NOGGINPLUS 250 SMARTCART

Reconstruction of older city areas often results in the burial of historic features. Foundations, roads and rail lines are

often covered over by new construction. Our latest Noggin case study comes from Atlanta, Georgia.

An archaeologist survey in 1994 discovered sections of buried rail tracks in a downtown Atlanta area (see map in Fig-ure 1) where Kelly Street had been disrupted by the construction of the I-20 freeway. The tracks are believed to be

trolley tracks constructed in 1899 by the Atlanta Railway Company for electric trains linking downtown Atlanta to

Grant Park which is now recognized as a Georgia historic district.

Figure 1: The GPR survey was carried out on a part of a relocated road that had been put over trolley tracks laid down in the late

1800’s.

The Kelly Street Trolley Tracks are now being considered for inclusion in the National Register of Historic Places.

The Georgia Department of Transportation (DOT) and area planners needed to relocate the buried tracks and map out

their extent. Selected excavation of parts of the area is being considered to evaluate the methods and materials used

in the early trolley line construction.

8/20/2019 GPR Annan 2003



260

Figure 2: The survey grid of 10m by 10m (right) clearly shows the trolley tracks when plotted using EKKO_Mapper software (left)

Before any excavation could be contemplated, a means to locate the buried tracks was required. While in the area, a

Sensors & Software Inc. staff member tested the usage of GPR for this site evaluation with a quick Noggin SmartCart

survey. An exploratory 10m x 10m grid was established in the area where the lines were previously seen. A Noggin- plus 250 reconnaissance survey was conducted with 0.5 m line spacing.

The resulting data were then processed to quickly create a map on site using EKKO_Mapper. The results are shown

in Figure 2. A depth slice centered at 0.25 m below surface clearly indicated two sets of trolley tracks curving around

the corner on the previous Kelly Street alignment. A tighter survey line spacing and higher GPR frequency are rec-

ommended for future surveys to improve spatial resolution.

Figure 3: An Atlanta trolley car typical of the early 1900’s.

Archaeologists, urban planners and construction engineers are converts to the power of Noggin SmartCart systems

with integrated GPS positioning and fast mapping output with EKKO_Mapper. The resulting subsurface imaging

solution cannot be matched.

8/20/2019 GPR Annan 2003



261

ON-SITE PROCESSING & 3-D IMAGING

New transitions are occurring with concrete inspection as demonstrated by the Sensors & Software Inc.’s Conquest.

Users measure data following a prescribed menu to image an area. The data are immediately turned into depth slice

maps on site in minutes as shown below. In this system the vector nature of the measurement is critical. Imaging

makes use of vector information to generate pseudoscalar depth slice maps.

Photo of Conquest GPR system designed specifically for imaging concrete. Data are acquired on a predefined survey

grid.

Example of depth slice output of a Conquest system showing rebar and an electrical conduit. Conquest generatesdepth slice maps of a concrete structure enabling the location of embedded objects such as rebar, post tension cables

and utility conduits. The slices show two depth images from the same location. The left sees a rebar grid and an elec-

trical conduit meeting at a junction box. The right shows another conduit at a greater depth.

8/20/2019 GPR Annan 2003



262

10.6 MILITARY, LAW ENFORCEMENT & ESPIONAGE

UXO DETECTION

8/20/2019 GPR Annan 2003



263

10.7 BIO APPLICATIONS

WOODEN POLE & TREE MAPPING

A unique but extremely effective use of ground penetrating radar (GPR) is mapping the internal structure of wooden poles and trees. GPR has proven to be a powerful means to locate anomalies in wood, primarily associated with

changes in water content. Variations in water content most often indicate that rot or insect infestation is present.

This survey was conducted in Hong Kong to determine whether or not rot was present in wooden hydro poles. The

study was easily carried out by one person using a pulseEKKO 1000 GPR system with 1200 MHz antennas to obtain

a high-definition image of the poles. This information is very important in ascertaining the stability of poles support-

ing telephone and hydro cabling.

The above data section clearly shows the rot present in the first 200 cm of the pole. The GPR information indicated

that the rot was extensive, making the pole unsuitable for further use.

8/20/2019 GPR Annan 2003



264

10.8 SNOW & ICE

GLACIOLOGY

Ground penetrating radar (GPR) is an ideal method for glaciology as ice is very transparent to GPR signals.

This study was conducted on Bylot Island, NWT, Canada, using a pulseEKKO GPR system with 50 MHz antennas.

In ice covered areas, knowledge of ice thickness and permafrost condition is important for land use management and

resource exploitation planning.

The data above were quickly and easily collected by a two man team, using sleds to expedite the process. The pul-

seEKKO GPR data above provide a graphic view of the ice conditions. Note the clear definition of the glacier base

and the cracks in the ice above.

Data compliments of Carleton University, Department of Geology, Canada

8/20/2019 GPR Annan 2003



265

WATER CONTENT IN SNOW

Measuring snow thickness is a fairly straightforward application for ground penetrating radar (GPR). Portable GPR

systems mounted on sleds, vehicles, or towed by skiers can quickly profile snow depth. GPR also gives users the

power to estimate water content of snow, making it an optimal tool for snow studies.

These data were obtained using a pulseEKKO 1000 system with 900 MHz antennas during a survey at a hydroelectric

power generation research area in Svalbard, Norway. In this case the water content of the snow was the information

required to estimate the availability of water for hydro power generation. The data collected distinctly show the snow

thickness and the location of buried objects. Water content was found by acquiring CMP velocity soundings through

the snow and combining these with the reflection profiles shown above, using the integrated pulseEKKO software.

The speed of snow pack mapping combined with the ability to estimate water content are used in hydroelectric flow

models, which help plan reservoir capacity for power generation. The ability to detect buried objects can be very use-

ful in avalanche areas, where time is of the essence when locating buried victims.

Data compliments of TS Geokonsult, Sweden

8/20/2019 GPR Annan 2003



266

SKI SLOP SNOW THICKNESS MAPPING

Historically, GPR practitioners have looked at GPR cross sections as the fundamental information. Some applica-tions no longer look at GPR sections but treat them as a given and only display derived data. This is a sign of increas-ing maturity and eventually one will seldom see raw GPR sections used for presentation or interpretation. More andmore derived products of various attributes (reflectivity impedance) and quantitative numbers (depth, target size, etc.)

will be extracted which are specific to a particular application.

A good example is the SnowScan snow thickness application. In the initial measurement process, the data were displayed as radar sections. Within months the application had auto picking and tracking capability which displayed asnow depth profile and could output the information as a single depth number versus position using GPS (global posi-tioning system) location. In its current embodiment the system no longer needs to display radar data (or any data for that matter) but logs snow thickness along with the GPS coordinates. These data are normally transformed into mapswithout ever looking at the raw GPR cross section.

Photos of a compact, lightweight SnowScan GPR designed specifically to measure snow depth and combine with GPS

positioning.

Example of a snow depth map created using a SnowScan GPR system.

8/20/2019 GPR Annan 2003



267

ICE ROADS & GPR

GPR for ice thickness measurements on glaciers is well known. Less well known is that measuring ice thickness for

roads is another GPR application. Ice roads are a common transportation link in northern regions of Canada, Alaska

and northern Europe. Quite often ice roads provide the only economical way to get materials to-and-from remote

mining communities and petroleum exploration sites.

The history of GPR ice road thickness goes back to the 1970’s. In the early days of oil and gas developments in

Alaska and in Canada’s Northwest Territories, GPR was used for an umber of applications, one of which was measur-

ing ice thickness. Sensors & Software’s staff was actively involved in pioneering activities which demonstrated the

viability of GPR for ice road evaluation in the Arctic.

For those not familiar with ice roads, these roads are constructed on natural bodies of water such as lakes, rivers, and

streams after the onset of natural bodies of freezing occurs in the winter. The major question, of course, is how safe

is the ice for a vehicle? Often the snow cover is plowed off the ice to encourage thickening of the ice so it has the

strength to carry larger trucks and other heavy vehicles.

In addition to roads, temporary aircraft runways are constructed on ice. As with the roads, it is important to know ice

thickness and to determine if degradation is occurring. Thinning ice occurs because of change in current patterns,warming weather conditions or ice movement.

Recently surges in oil and gas prices have stimulated exploration and development in northern areas again. As a

result, this has stirred interest in GPR for construction of safe ice roads and monitoring their condition during winter

operations.

Figure 1: GPR mounted behind a truck on the Mackenzie River delta near Inuvik, Canada.

Figure 1 shows a recent installation on a truck. This system was mounted on a vehicle for regular evaluation of ice

thickness on an ice road in the Mackenzie River delta. In this case, the sensor is towed behind a pick-up truck; some-

times a lighter vehicle such as skidoo is used. In some risky applications, the operators prefer to push the transducer

ahead of the vehicle rather than trail it behind for obvious safety reasons.

8/20/2019 GPR Annan 2003



268

Figure 2: The ice thickness monitor mounted in the cab of the pick-up truck gives the operator a real time read-out of ice thick-

ness.

Figure 2 shows the operation and control part of the system which provides a monitor of ice thickness in real time.

The display is mounted in the truck cab right beside the driver so a view of ice thickness is available at all times.

Positioning of the data can be done in a number of manners. One can use an odometer wheel, landmarks or a GPS

positioning system to get coordinates. The particular configuration is dependent on the particular needs of the appli-

cation. The Sensors & Software Inc. DVL (digital video logger) is designated to readily accommodate all options.

Figure 3: Example profile of ice thickness on a 1 km stretch of the Mackenzie River delta obtained in December 2000.

Figure 3 presents a sample profile of data from the Mackenzie River delta. Ice at this site was about 30 to 70 cm

thick. A 500 MHz bandwidth GPR transducer was used for this particular survey operation. The data were acquired

at about 30 km/hour.

This novel use of GPR is just another example of the breadth of GPR applications. The fascinating application of

GPR will no doubt stimulate other applications for the technology in snow and ice as imaginations are put to work.

8/20/2019 GPR Annan 2003



269

10.9 EDUCATIONAL MATERIAL

GPR FREQUENCY SELECTION

A major factor in the success of your ground penetrating radar (GPR) survey is the use of the correct antenna fre-

quency. Defining a goal before you start will help ensure that you get the desired results. This case history demon-

strates the effects of different antenna frequencies used on the same site.

The pulseEKKO GPR data below were collected over two highway tunnels in Sweden. The first data set was col-

lected with 100 MHz antennas. The data indicate the two road tunnels cut through the bedrock, and also shows bed-

rock stratigraphy. Note that the tunnel reflectors are weak and the gneissic bedrock texture events are strong.

The second data set, collected with 50 MHz antennas, show much stronger returns from the two tunnels. Note that

these data more readily locate a culvert and a communications cable to the right of the tunnels, shown by arches on

the right. Note that this data set is much less cluttered by the bedrock texture.

This is one example of how proper antenna frequency can effect the outcome of your survey. If the objective is large

scale structure mapping, the 50 MHz data nicely define underground structures with minimal bedrock feature inter-ference. On the other hand, if the purpose is to map mass stability, for tunnel maintenance and repair, the 100 MHz

data will provide more information of potentially hazardous bedrock conditions.

8/20/2019 GPR Annan 2003



270

DATA MIGRATION

Once ground penetrating radar (GPR) profiles are obtained, migration of the image data can reconstruct the appear-

ance of targets to more closely resemble the original geometry. The ability to display data in different forms helps in

the interpretation of data.

The data below were acquired at a test site with buried pipes and barrels. The raw data in Figure 1 are what you

would see without any migration. The pipes and barrels exhibit classic hyperbolic responses which are indicative of

localized types of targets.

The data shown in Figure 2 is result of applying an F-K migration algorithm to the unmigrated data set assuming a

constant background velocity. This migrated data shows a more correct representation of the ground.

The data displayed in Figure 3 are further processing of the migrated data set (Figure 2), which give a fuzzy image of

the actual targets. The small pipes at nine and 14 meters are sharpened to point-like events. The 0.5 meter diameter

pipe at 4 meters and the barrels at 19 and 24 meters remain extended in space.

Note that while the different displays of data are optimal for interpretation, most experienced GPR users find the clas-

sic hyperbolic shapes as shown in the first data set more readily identifiable.

8/20/2019 GPR Annan 2003


GPR Principles, Procedures & Applications 11-References

271

11 R EFERENCES

Alumbaugh, D.L. and Newman, G.A., 1994, Fast Frequency-Domain Electromagnetic Modeling of a 3-D EarthUsing Finite Differences: Extended Abstracts from the Society of Exploration Geophysicists 1994 Annual Meeting,Los Angeles, California, pp. 369-373.

Annan, A.P., 1974, The Equivalent Source Method for Electromagnetic Scattering, Ph.D. Thesis, Memorial Uni-versity.

Annan, A.P. and Chua, L.T., 1992, Ground Penetrating Radar Performance Predictions: The Geological Surveyof Canada, Paper 90-4, pp. 5-13.

Annan, A.P. and Davis, J.L., 1976, Impulse Radar Soundings in Permafrost: Radio Science, Vol. 11, pp. 383-394.

Annan, A.P. and Davis, J.L., 1977, Radar range analysis for geological material; Report of Activities, Part B, Geo-logical Survey of Canada, Paper 77-1B, p. 117-124.

Annan, A.P. and Davis, J.L., 1978, Methodology for Radar Transillumination Experiments: Report of Activities,Geological Survey of Canada, Paper, 78-1B, pp. 107-110.

Annan, A.P., Davis, J.L., and Gendzwill, D., 1988, Radar Sounding in Potash Mines: Saskatchewan, Canada. Geophysics, Vol. 53, pp. 1556-1564.

Annan, A.P. and Cosway, S.W., 1992, Ground Penetrating Radar Survey Design: Proceedings of the Symposiumon the Application of Geophysics to Engineering and Environmental Problems, SAGEEP’92, April 26-29, 1992,Oakbrook, IL, pp. 329-351.

Annan, A.P. and Cosway, S.W., 1994, GPR Frequency Selection: Proceedings of the Fifth International Conferenceon Ground-Penetrating Radar, Kitchener, Ontario, Canada, June 12-16, 1994, pp. 747-760.

Annan, A.P. and Cosway, S.W., 1992, Simplified GPR Beam Model for Survey Design, Extended abstract of 62ndAnnual International Meeting of the Society of Exploration Geophysicists, New Orleans, 1992.

Annan, A.P., 1993, Practical Processing of GPR Data: Proceedings of the Second Government Workshop on

Ground Penetrating Radar, October, 1993, Columbus, Ohio.

Annan, A.P., Davis, J.L. and Johnston, G.B., 1997, Maximizing 3D GPR Image Resolution: A Simple Approach: Proceedings of the High Resolution Geophysics Workshop, University of Arizona, Tucson, AZ, January 6-9, 1997.

Annan, A. P., 1996, Transmission Dispersion and GPR, JEEG, Vol. 0, January 1996, p. 125-136.

Annan, A.P., 1973, Radio Interferometry Depth Sounding: Part I - Theoretical Discussion: Geophysics, Vol. 38, p. 557-580.

Annan, A.P., Waller, W.M., Strangway, D.W., Rossiter, J.R., Redman, J.D., and Watts, R.D., 1975, The Electromag-netic Response of a Low-Loss, 2-Layer Dielectric Earth for Horizontal Electric Dipole Excitation: GeophysicsV. 40 p. 285-298.

Annan, A.P., Redman, J.D., Pilon, J.A., Gilson, E.W., and Johnston, G.B., 1997, Crosshole GPR for Engineeringand Environmental Applications. Proceedings of the High Resolution Geophysics Workshop, University of Ari-zona, Tuscon, AZ, January 6-9, 1997.

Bailey, J.T., Evans, S. and Robin G. de Q., 1964, Radio echo sounding in polar ice sheets: Nature, Vol. 204, No.4957, p. 420-421.

Banos,A., 1962, Dipole Radiation in the Presence of a Conducting Half-Space, Pergamon Press.

Barakat, N., Dolphin, L., et al., 1975, Electromagnetic Sounder Experiments at the Pyramid of Giza: FinalReport, NSF Grant No. GF-38767, Stanford Research Institute, Menlo Park, California.

8/20/2019 GPR Annan 2003


11-References GPR Principles, Procedures & Applications

272

Benson, R.C., Glaccum, R.A. and Noel, M.R., 1984, Geophysical Techniques for Sensing Buried Wastes andWaste Migration. US EPA Contract No. 68-03-3053. Environmental Monitoring Systems Laboratory. Office ofR&D. US EPA, Las Vegas, Nevada 89114. 236p.

Bentley, C.R., 1964, The structure of Antarctica and its ice cover: Research in Geophysics, Vol. 2: Solid Earth andInterface Phenomena, Cambridge Mass., Technology Press of Massachusetts Institute of Technology, p. 335-389.

Bergmann, T., Blanch, J.O., Robertsson, J.O.A., Holliger, K., 1999, A simplified Lax-Wendroff correction forstaggered-grid FDTD modeling of electromagnetic wave propagation in frequency-dependent media, Geophys-ics, Vol. 64, pp. 1369-1377.

Bergmann, T., Robertsson, J.O.A., and Holliger, K., 1996, Numerical properties of staggered finite-differencesolutions of Maxwell’s equations for ground-penetrating radar modeling, Geophys. Res. Lett., 23, pp. 45-48.

Berkhout, A. J., 1984, Seismic Resolution: Resolving Power of Acoustical Echo Techniques. Geophys. Press.

Bevan, V. and Kenyon, J., 1975, Ground-penetrating radar for historical archaeology: MASCA Newsletter, Vol.II, No. 2, University of Pennsylvania Museum of Applied Science Center for Archaeology, Philadelphia, Pennsylva-nia.

Bleistein, N. and Gray, S.H., 2001, From the Hagedoorn imaging technique to Kirchhoff migration and inver-sion, Geophysical Prospecting, Vol. 49, pp. 629-643.

Born, M., & Wolf, E., 1980, Principles of Optics. 6th Edition, Pergamon Press.

Brekhovskikh, L.M., 1960, Waves in Layered Media, New York, Academic Press.

Brewster, M.L., 1993, Observed migration of a controlled DNAPL release by ground penetrating radar.M.Sc. Thesis, University of Waterloo.

Brewster, M.L. and Annan, A.P., 1994, Ground-penetrating radar monitoring of a controlled DNAPLrelease: 200 MHz radar: Geophysics, Vol. 59, P. 1211-1221.

Burhl, M., Vermeer, G.J.O., Kiehn, M., 1996, Fresnel Zones for Boradband Data, Geophysics, 61, No. 2, p 600 -604.

Cai, J. and McMechan, G.A., 1995, Ray-based Synthesis of Bistatic Ground-Penetrating Radar Profiles, Geo- physics, Vol. 60, pp. 87-96.

Carcione, J.M., 1996, Ground-penetrating radar: Wave theory and numerical simulation in lossy anisotropicmedia, Geophysics, Vol. 61, p. 1664-1677.

Casper, D.A. and Kung, K-J.S., 1996, Simulation of Ground-Penetrating Radar Waves in a 2-D Soil Model, Geo- physics, Vol. 61, pp. 1034-1049.

Cook, J.C., 1973, Radar Exploration Through Rock in Advance of Mining, Trans. Society Mining Engineers,AIME, V. 254, pp. 140-146.

Coon, J.B., Fowler, J.C. and Schafers, C.J., 1981, Experimental Uses of Short Pulse Radar in Coal Seams, Geo- physics, Vol. 46, No. 8, pp. 1163-1168.

Cerveny, V. and Ravindra, R., 1971, Theory of Seismic head waves, University of Toronto, Press, pp. 312

Davis, J.L., and Annan, A.P., 1989, Ground Penetrating Radar for High-Resolution Mapping of Soil and RockStratigraphy, Geophysical Prospecting, 37 p. 531-551.

Davis, J.L., and Annan, A.P., 1986, Borehole Radar Sounding in CR-6, CR-7 and CR-8 at Chalk River, Ontario,Technical Record TR-401, Atomic Energy of Canada Ltd.

8/20/2019 GPR Annan 2003



273

Davis, J.L., and Annan, A.P., 1989, Ground Penetrating Radar for High-Resolution Mapping of Soil and RockStratigraphy, Geophysical Prospecting, Vol. 37, pp. 531-551.

Davis, J.L., Annan, A.P., Black, G., and Leggatt, C.D., 1985, Geological sounding with low frequency radar: inExtended Abstracts, 55th Annual International Meeting of the Society of Exploration Geophysicists, Washington,D.C.

Davis, J.L., Annan, A.P., and Vaughan, C.J., 1985, Placer Exploration Using Radar and Seismic Methods, Cana-dian Institute of Mining Bulletin 80, (898), pp. 67-72.

Davis, J.L., Heginbottom, J.A., Annan, A.P., Duncan, K.E., Plan-View Presentation of GPR Data, Proceedings ofthe Seventh International Conference on Ground Penetrating Radar (GPR ‘98), University of Kansas, Lawrence, KS

Dix, C.H., 1956, Seismic Prospecting of Oil, Harper, New York.

Dolphin, L.T., et al., 1978, Radar Probing of Victorio Peak, New Mexico: Geophysics, Vol. 43, No. 7, p. 1441-1448.

Doolittle, J.A., and Asmussen, L.E., 1992, Ten years of applications of Ground Penetrating Radar by UnitedStates Department of Agriculture, Proceedings of the Fourth International Conference on Ground PenetratingRadar, Geological Survey of Finland, Special Paper 16, p. 139-147.

El Said, M.A.H., 1956, Geophysical prospection of underground water in the desert by means of electromag-netic interference fringes, Pro. I.R.E., v. 44, p. 24-30 and 940.

Endres, A. L. and Knight, R., 1992, A theoretical treatment of the effect of microscopic fluid distribution on thedielectric properties of partially saturated rocks, Geophysics Prospector, vol. 40, p. 307 - 324

Evans, S., 1965, Dielectric properties of ice and snow - a review, Journal of Glaciology, Vol. 5, No. 42, p. 773-792.

Fisher, E., McMechan, G.A., and Annan, A.P., 1992, Acquisition and processing of wide-aperture ground pene-trating radar data: Geophysics, 57, p. 495.

Fisher, E., McMechan, G.A., Annan, A.P., and Cosway, S.W., 1992, Examples of reverse-time migration ofsingle-channel, ground-penetrating radar profiles, Geophysics, 57, p. 577-586.

Friere, S.L.M., and Ulrych, T.J., 1988, Application of singular decomposition to vertical seismic profiling,Geophysics, v. 53, n. 6, p. 778-785.

Gerlitz, K., Knoll, M.D., Cross, G.M., Luzitano, R.D., and Knight, R., 1993, Processing ground penetratingradar data to improve resolution of near-surface targets, Proceeding of the Symposium on the Application ofGeophysics to Engineering and Environmental Problems, San Diego, California.

Giannopoulos, A., 1997, The investigation of Transmission Line Matrix and Finite-Difference Time-DomainMethods for the Forward Problem of Ground Probing Radar, D. Phil thesis, Department of Electronics, Univer-sity of York, UK.

Gilson, E.W., Redman, J.D., Pilon, J.A., and Annan, A.P., 1996, Near Surface Applications of Borehole Radar.Proceedings of the Symposium on the Application of Geophysics to Engineering and Environmental Problems, April28 - May 2, 1996, Keystone, Colorado, pp. 545-553.

Goodman, D., 1994, Ground-Penetrating Radar Simulation in Engineering and Archaeology, Geophysics, Vol.59, pp. 224-232.

Grant, F.S. and West, G.F., 1965, Interpretation Theory in Applied Geophysics, McGraw-Hill.

Grasmueck, M., 1996, 3-D Ground-Penetrating Radar Applied to Fracture Imaging in Gneiss, Geophysics, Vol.61, pp. 1050-1064.

Greaves, R.J., Lesmes, D.P., Lee, J.M. and Toksoz, M.N, Velocity Variation and Water Content Estimated fromMulti-Offset, Ground Penetrating Radar, Geophysics, Vol. 61, No. 3, May-June 1996, pp. 683-695.

8/20/2019 GPR Annan 2003



274

Greaves, R.J. and Toksoz, M.N., 1994, Applications of Multi-Offset, Ground-Penetrating Radar, Proceedings ofthe Symposium on the Application of Geophysics to Engineering and Environmental Problems, SAGEEP’94, pp.775-793.

Green, A., Holliger, K., Horstmeyer, H., Maurer, H., Tronicke, J., van der Kruk, J., 2002, 3D acquisition, processingand imaging of Ground Penetrating Radar Data, GPR 2002 Tutorial 2 Notes, Proceedings of the Ninth Interna-tional Conference on Ground Penetrating Radar (GPR 2002), Santa Barbara, California, April 29 – May 2, 2002.

Hansen, S., 1942, Electrical Wave Analysis, US Patent 2,280,524.

Hasted, J. B., 1972, Liquid water -dielectric properties: in Water a comprehensive treatise, v. 1, The physics and physical chemistry of water, F. Franks, ed., Plenum Press, NY, p.255-310

Hohmann, 1987, Numerical Modelling for Electromagnetic Methods of Geophysics, Electromagnetic Methods inApplied Geophysics – theory: Volume 1, edited by M.N. Nabighian, Society of Exploration Geophysicists, pp. 313-361.

Holliger, K. and Bergmann T., 2000, Finite-difference modelling of borehole georadar data, in press, Geophysics.

Holser, W.T., Brown, R.J.S., Roberts, F.A., Fredrikkson, O.A., and Unterberger, R.R., 1972, Radar logging of a saltdome: Geophysics, 37, P. 889-906.

Holtzman, R. and Kastner, R., 2001, The Time-Domain Discrete Green’s Function Method (GFM) characteriz-ing the FDTD Grid Boundary, IEEE Transactions of Antennas and Propagation, Vol. 49, No. 7, July 2001, pp.1079-1093.

Jackson, J.D., 1962, Classical Electrodynamics, John Wiley and Sons.

Jiracek, G.R., 1967, Radio Sounding of Antarctic Ice: Research Report Series 67-1, The University of Wisconsin,Geophysical & Polar Research Center, Department of Geology.

Jol, H., 1996, Digital Ground Penetrating Radar (GPR): A New Geophysical Tool for Coastal Barrier Research(Examples from the Atlantic, Gulf and Pacific Coasts U.S.A.), Journal of Coastal Research, Fall 1996.

Kline, M. and Kay, I.W., 1965, Electromagnetic Theory and Geometrical Optics, Interscience Publishers, NewYork.

Knapp, R. W., 1991, Fresnel Zones in the Light of Broadband Data: Geophysics, 56, P. 354 - 359.

Lampe, B. and Holliger, K., 2000, Finite-difference modelling of ground-penetrating radar antenna radiation,556 bis 560, Proceedings of the 8th International Conference on Ground Penetrating Radar, Gold Coast, Australia,May 23-26, 2000.

Luneburg, R.K., 1964, Mathematical Theory of Optics, University of California, Press, Berkeley.

Maijala, P., 1992, Application of Some Seismic Data Processing Methods to Ground Penetrating Radar Data,Fourth International Conference on Ground Penetrating Radar June 8-13, 1992, Rovaniemi, Finland. Geological Sur-vey of Finland, Special Paper 16, 365 pages.

McMechan, G.A., 1981, Modeling of Zero-offset Profiles with Asymptotic Ray Theory: Can J. Earth Sci., vol. 18,

p. 551-560.

Mooney, H. M. and Bleifuss, R., 1953, Magnetic Susceptibility Measurements in Minnesota, Part II, Analysis offield results, Geophysics, vol. 18, p 383-393.

Moran, M.L. and Arcone, S.A., 2000, GPR radiation pattern effects on 3D Kirchhoff Imaging, Proceedings of the

8th International Conference on Ground Penetrating Radar, Gold Coast, Australia, May 23-26, 2000.

8/20/2019 GPR Annan 2003



275

Morey, R.M., 1974, Continuous Subsurface Profiling by Impulse Radar: Proceedings of Engineering FoundationsConference on Subsurface Exploration for Underground Excavations and Heavy Construction, Henniker, N.H. p.213-232.

Neuf, D., Brown, D., and Jaracz, R., 1973, Multioctave Double Balanced Mixer: Microwave Journal, Vol. 16, No. 1(Jan. 1973), pp. 13-14,

Noon, D.A., Longstaff, D., Yelf, R.J., 1994, Advances in the development of step frequency ground penetratingradar; Fifth International Conference on Ground Penetrating Radar (GPR ‘94), Kitchener, ON, Canada, June 12-16,1994, Vol. 1 of 3, p. 117-131.

Oguz, U. and Levent, G., 2001, Modelling of Ground-Penetrating-Radar Antennas with Shields and SimulatedAbsorbers, IEEE Transactions on Antenna and Propagation, Vol. 49, No. 11, November 2001, pp. 1560-1567.

Olhoeft, G.R., 1975, The Electrical Properties of Permafrost, Ph.D. Thesis, University of Toronto, 172 pages.

Olhoeft, G.R., 1981, Electrical Properties of Rock and Minerals Short Course Notes, Pers. Com., 196 pages.

Olhoeft, G.R., 1987, Electrical Properties from 10-3 to 10+9 Hz - Physics and Chemistry, Proceedings of the 2nd International Symposium on the Physics and Chemistry of Porous Media, American Institute of Physics ConferenceProceedings, Vol. 154, pp. 281-298.

Olhoeft, G.R., 1988, Interpretation of Hole-to-Hole Radar Measurements, Proceedings of the Third TechnicalSymposium on Tunnel Detection, January 12-15, 1988, Golden, CO, pp. 616-629.

Oldenburg, D.W., McGillivary, P.R., and Ellis, R.G., 1993, Generalized subspace method for large scale inverseproblems, Geophys. J. Internat., Vol. 144, pp. 12-20.

Olsson, O., Falk, L., Forslund, O., and Sandberg, E., 1987, Crosshole Investigations - Results from BoreholeRadar Investigations, Stripa Project TR 87-11. SKB, Stockholm, Sweden.

Olsen, O., Anderson, P., Carlsten, S., Falk, L., Niva, D., and Sandberg, E., 1992. Fracture Characterization inCrystalline Rock by Borehole Radar, Ground Penetrating Radar, Geological Survey of Canada, Paper 90-4, pp.139-150.

Owen, T.R., 1981, Cavity Detection Using VHF Hole to Hole Electromagnetic Techniques: Proceedings of theSecond Tunnel Detection Symposium, Colorado School of Mines, Golden CO, July 21-23, 1981, U.S. Army MERA-DOM, Ft. Belvoir, VA, pp. 126-141.

Ott, H., 1941, Reflexion and Brechung von Kugelwellen: Effekte Q. Ordnung: Ann. Physik, v. 41, p. 443-466

Powers, M.H. and Olhoeft, G.R, 1994, Modeling Dispersive Ground Penetrating Radar Data: Proceedings of theFifth International Conference on Ground-Penetrating Radar, Kitchener, Ontario, Canada, June 12-16, 1994.

Redman, J.D., Kunert, Gilson, E.W., M., Pilon, J.A., Annan, A.P., 1996, Borehole Radar for Environmental Appli-cations: Selected Case Studies: Proceedings of the Sixth International Conference on Ground Penetrating Radar(GPR’96), September 30-October 3,1996, Sendai, Japan.

Rees, H.V., and Glover, J.M., 1992, Digital enhancement of ground probing radar data, in ground penetratingradar, ed. J. Pilon; Geological Survey of Canada, Paper 90-4, p. 187-192.

Reitz, J.R. and F.J. Milford, 1960, Foundations of Electromagnetic Theory, Addison-Wesley.

Roberts, R.L. and Daniels, J.J., 1996, Analysis of GPR Polarization Phenomena, JEEG, vol. 1, no. 2, p. 139-157.

Roddy, D., 1992, Cepstral analysis of subsurface data, in ground penetrating radar, ed. J. Pilon; Geological Sur-vey of Canada, Paper 90-4, p. 193-198.

Rogers, P.G., Edwards, S.A., Young, J.A., and Downey, M., 1987, Geotomography for the Delineation of CoalSeam Structure, Geoexploration, Vol. 24, pp. 301-328.

8/20/2019 GPR Annan 2003



276

Sanders, K.A., 1994, Characterization of DNAPL Movement in Saturated Porous Media Using Ground Pene-trating Radar, M.E. Thesis ER-4336, Dept. of Geology, Colorado School of Mines, Golden, CO. 258.

Scaife, J.E., & Annan, A.P., 1991, Ground Penetrating Radar - A Powerful, High Resolution Tool for MiningEngineering and Environmental Problems; Paper presented at 93rd CIM Annual General Meeting, Vancouver,B.C., April 29 - May 1, 1991.

Schmitt, H.J., Harrison, C.W., Williams, C.S., 1996, Calculated and Experimental Response of Thin CyndricalAntennas to Pulse Excitation; IEEE Trans. on Antennas & Propagation, AP14, p. 120-126.

Sigurdsson, T., 1993, Ground Penetrating Radar for Geological Mapping, Thesis, Aarhus University, Departmentof Earth Sciences, Aarhus, Denmark, August, 1993.

Sigurdsson, T. and Overgaard, T., 1996, Application of GPR for 3D Visualization of Geological and StructuralVariation in a Limestone Formation, Proceedings of the Sixth International Conference on Ground PenetratingRadar (GPR’96), September 30-October 3,1996, Sendai, Japan.

Simmons, G., Strangway, D., Annan, A.P., Baker, R., Bannister, L., Brown, R., Cooper, W., Cubley, D., deBetten-court, J., England, A.W., Groener, J., Kong, J.A., LaTorraca, G., Meyer, J., Nanda, V., Redman, J.D., Rossiter, J.,Tsang, L., Urner, J., Watts, R., 1973, Surface Electrical Properties Experiment, in Apollo 17: Preliminary ScienceReport, Scientific and Technical Office, NASA, Washington, D.C., pp. 15-1 - 15-4.

Smith, G.S., 198, Directive Properties of Antennas for Transmission into a Material Half-Space, IEEE Trans.Antennas propagate, vol AP-32, p. 232-246.

Smythe, W.R., 1989, Static & Dynamic Electricity, Taylor & Francis, A SUMMA book, 1989.

Sommerfeld, A., 1949, Partial Differential Equations in Physics, New York, Academic Press.

Stewart, R.D. and Unterberger, R.R., 1976, Seeing through rock salt with radar: Geophysics, 41, P. 123-132.

Stickley, G.F., Noon, D.A., Cherniakov, M., Longstaff, I.D., Leat, C.J., Li, C.W., 1998, Gated Stepped-FrequencyGPR Field Demonstrations, Seventh International Conference on Ground-Penetrating Radar (GPR ‘98), Universityof Kansas, Lawrence, Kansas, USA, May 27-30, 1998, Volume 1, p. 343-347.

Tarantolo, P.J., and Unterberger, R.R., 1978, Radar detection of boreholes in advanced mining: Geophys Prosp.

26, P. 359-382.

Tarantota, A., 1987, Inverse Problem Theory, Elsevier Science E.V., Amsterdam

Telford, W.M., Geldart, L.P., Sheriff, R.E., Keys, D.A., 1976, Applied Geophysics, Cambridge University Press.

Theirbach, R., 1974: Electromagnetic Reflections in Salt Deposits: J. Geophys. 40, P. 633-637.

Tillard, T., and Dubois, J-C., 1992, Influence and lithology on radar echoes: analysis with respect to electromag-netic parameters and rock anisotropy: Fourth International Conference on Ground Penetrating Radar June 8-13,1992. Rovaniemi, Finland. Geological Survey of Finland, Special Paper 16, 365 pages.

Todoeschuck, J.P., Lafleche, P.T., Jensen, O.G., Judge, A.S., and Pilon, J.A., 1992, Deconvolution of ground prob-ing radar data, in ground penetrating radar, ed. J. Pilon; Geological Survey of Canada, Paper 90-4, p. 227-230.

Turner, G., 1992, Propagation deconvolution, Fourth International Conference on Ground Penetrating Radar June 8-13, 1992. Rovaniemi, Finland. Geological Survey of Finland, Special Paper 16, 365 pages.

Topp, G.C., Davis, J.L. and Annan, A.P., 1980, Electromagnetic Determination of Soil Water Content: Measure-ments in Coaxial Transmission Lines, Water Resources Research, Vol. 16, No. 3, p. 574-582.

Ulriksen, C.P.F., 1982, Application of Impulse Radar to Civil Engineering, Unpublished Ph.D. Thesis, Dept. ofEngr, Geol., U. of Technology, Lund, Sweden, p. 175.

8/20/2019 GPR Annan 2003



277

Unterberger, R.R., 1978, Radar propagation in rock salt: Geophys. Prosp., 26, P. 312-328.

van der Kruk, J., 2001, Three dimensional imaging of multi-component ground penetrating radar, Ph.D. Thesis,Delft University of Technology, pp. 242.

Vickers, R.S., 1978, Application of Resistivity and Radar Techniques to Archaeological Surveys: SRI TechnicalMemorandum.

Vickers, R.S. and Bollen, R.S., 1974, An Experiment in the Radio Echo Sounding of Temperate Glaciers: FinalReport, Contract 14-08-0001-14650 for U.X. Geological Survey, Denver, Stanford Research Institute, Menlo Park,California.

Vozoff, K., Smith, G.H., Hatherly, P.J., and Thompson, S., 1993, An Overview of the Radio Imaging Method inAustralian Coal Mining, First Break, Vol. 10, (January 1993), pp. 13-21.

Von Hippel, A.R., ed., 1954, Dielectric Materials and Applications, John Wiley and Sons, New York, 438 p.

Wait, J.R., 1962, Electromagnetic Waves in Stratified Media, Pergamon Press, (Revised edition, 1970).

Waite, A.H. and Schmidt, S.J., 1961, Gross errors in height indication from pulsed radar altimeters operatingover thick ice or snow, IRE International Convention Record, Part 5, p. 38-54.

Walford, M.E.R., 1964, Radio echo sounding through an ice shelf : Nature, Vol. 204, No. 4956, p. 317-319.

Wang, T. and Tripp, A.C. 1996, FDTD Simulations of EM Waves Propagation in a 3-D Media, Geophysics, Vol.61, pp. 110-120.

Ward, S.H., Phillips, R.J., Adams, G.F., Brown, Jr., W.E., Eggleton, R.E., Jackson, P., Jordan, R., Linlor, W.I., Pee- ples, W.J., Porcello, L.J., Ryu, J., Schaber, G., Sill, W.R., Thompson, S.H. and Zelenka, J.S., 1972, Apollo LunarSounder Experiment, in Apollo 17: Preliminary Science Report, Scientific and Technical Office, NASA, Washing-ton, D.C., p. 22-1 - 22-26.

Watts, R.D., and England, A.W., 1976, Radio-echo Sounding of Temperate Glaciers: Ice Properties and SounderDesign Criteria. Journal of Glaciology, Vol. 21, No. 85, pp. 39-48.

Wedepohl, E., 1993, Radio Wave Tomography - Imaging Ore Bodies Using Radio Waves, Expanded Abstracts,

3rd Technical Meeting of the South African Geophysical Association, Capetown South Africa, pp. 85-88.

Wharton, R. P., Hazen, G. A., Rau, R. N., and Best, D. L., 1980, Advancements in electromagnetic propagationlogging: Society Petroleum Engineering, Paper 9041.

White, R.E., 1991, Properties of instantaneous seismic attributes, Geophysics: The Leading Edge of Exploration.

Wills, R.H., 1992, A digital phase coded ground probing radar; Ground penetrating radar, ed. J. Pilon; GeologicalSurvey of Canada, Paper 90-4, p. 231-235.

Wright, D.L., Bradley, J.A., Grover, T.P., 1994, Data acquisition systems for ground penetrating radar withexample applications from the air, the surface and boreholes; Fifth International Conference on Ground Penetrat-ing Radar (GPR ‘94), Kitchener, ON, Canada, June 12-16, 1994, Vol. 3 of 3, p. 1075-1089.

Xu, T., and McMechan, G.A., 1977, GPR attenuation and its numerical simulation in 2.5-dimensions, Geophys-ics, Vol. 62, pp. 403-414.

Yee, K.S., 1966, Numerical solutions of initial boundary value problems involving Maxwell’s equations in isoto-pic media, IEEE Trans. Ant. Propag., Vol. 14, pp. 302-307.

Yilmaz, O., 1987, Seismic Data Processing Series: Investigations in Geophysics, Vol. 2, Society of ExplorationGeophysicists, Tulsa.

8/20/2019 GPR Annan 2003



Young, J.L. and Nelson, R.O., 2001, A Summary and Systematic Analysis of FDTD Algorithms for Linearly Dis-persive Media, IEEE Antennas and Propagation Magazine, Vol. 43, No. 1, February 2001, pp. 61-77.

Documents

GPR Annan 2003